Before jumping on the ML/AI bandwagon of WebRTC media quality, make sure you’ve exhausted all of your other optimization alternatives.
TL;DR – make sure you optimize for media quality without AI before jumping to using AI…
In 2018 and 2019 at Kranky Geek we’ve started looking at machine learning. We’ve handpicked speakers and sessions who deal with these topics. We’ve done so for both voice and video technologies. The intent and idea behind this was to fit to the times. Everyone’s been doing AI so why not us in the context and domain of WebRTC and communication technologies?
It made perfect sense.
Then came 2020 and… changed everything. No one was really interested in AI or how to improve quality of experience with it. It was now used mainly for bots with the purpose of handling large loads of calls (call deflection and agent assist type technologies).
At times, it seemed like we were all back to basics. We now had to start scratching our heads and see what can be done to improve quality.Time for some quick wins
At Google and elsewhere, I am sure that a manager somewhere higher up came, saw the work that is being done, received an explanation how research into this machine learning stuff was progressing and showing promise, but in many ways required, well, more research, before it could be seen as anything that is close to being production ready.
And as managers do in these situations, they smack the table and say something like “I want quick wins”. So the developers went back to the basics. Trying to figure out what quick wins they can find to squeeze a bit more quality of that thing they had called WebRTC.
Quite surprisingly – it worked!
There seems to be ample room for optimizations. If you ask me? Someone forgot to try and squeeze this lemon properly.There’s more room for optimizations of WebRTC before we resort to machine learning Google’s optimizations of WebRTC’s code
It started somewhere with the pandemic.
One of the first indications was this tweet by Serge Lachapelle (former product manager for WebRTC at Google and leading Google Meet at the time of tweeting).April 17, 2020
Apparently, the video compositor wasn’t making the most out of the hardware it was using…
Since then we’ve seen some additional optimizations, though most of them taking place in the application level on top of the WebRTC implementation itself.
At Kranky Geek, Google discussed at length the optimization work it is working on. Mostly, making sure that video processing doesn’t take up too much CPU.Too many media format conversions in the WebRTC media pipeline
Apparently, Chrome is doing way too many video format conversions between getting the frames from the camera until it encodes and sends it out. Each conversion eats up CPU and I/O, generally killing the whole internal bus of the machine. Oh – and it means memory copies. Lots and lots of memory copies.
Video processing 101: zero copy is what you’re striving for.
We’re 10 years into WebRTC and the leading team behind WebRTC is just now starting to look at zero copying.
There are other areas and aspects where optimizations are taking place. Once the Kranky Geek videos will be ready and published, I’ll add the relevant one here.
Still got optimization juice in this lemon. Expect better performing WebRTC in the coming Chrome releases.Rushing towards 49-gallery view and 50+ group sizes
As the pandemic hit, Zoom grew. The media was filled with their gallery view.Zoom’s 49-gallery view. The holy grail of video group calls?
One use case that didn’t exist before the pandemic is large video calls. Up until today, we used to take these video meetings in the office inside meeting rooms. Cramming a few people in each room in a remote office and doing a call with 2-4 such rooms. Maybe someone joined from home or a hotel. You could see meetings with 10 participants. Sometimes. But the need just wasn’t really there.
The pandemic hit. People are now at home. And communicate with video remotely. A meeting of 4 became a meeting of 20 just because the participants are now sitting at home.
Even worse, schools are now remote. Each class has 20-40 students in it. And the teacher wants to see them all.
This made Zoom’s gallery view so popular (even if a tad useless if you ask me). It also made the magical number 49 magical. The holy grail of what is needed of a video conferencing service in a pandemic. Doesn’t matter if everyone is muting their video.
Microsoft and Google announced plans for supporting it. Then started running towards that value, each rising in the number of tiles in his gallery, reaching 49 recently.
Facebook grew from a meeting of 8 to meetings of 50.
Meetings are larger and longer now.
And again, we found the ways to make it happen with WebRTC.Best practices on group video scaling being rewritten
There are a lot of mechanisms in WebRTC that enable an application to squeeze the lemon and gain back CPU cycles as it tries to optimize for larger group calls.
But we never did have a place where all these are found and explained. A body of knowledge and understanding of how to make it happen.The larger the conference call size in WebRTC, the more complex the solution is going to be to implement it
In my recent/upcoming update to the Advanced WebRTC Architecture course there’s a lesson dedicated to this specific topic. It isn’t as if the information isn’t there in the course – it is spread all over the course. But now there’s a lesson on this alone. Because it became interesting only in 2020.
We have traded the focus on what is important to us with video communications. A video conference’s scale trumps quality at the moment. While I do understand we all want both all the time, but there is still a tradeoff between these two qualities of a system.The role of machine learning and AI in communications
Where does one fit machine learning and AI in this brave new world of large video conference calls?
Machine learning requires memory and CPU. Things we don’t have to spare at the moment in these large group calls. So we can’t just slap machine learning inference algorithms on the edge inside the web browser easily.
Edge inference in web browsers using WebAssembly is also brand new. So there’s no guide book to work with.
We won’t be using it to improve video quality or audio quality in the edge – we can’t really. Not enough CPU to spare.
There’s no real place for it on the server side either – that one requires decoding and encoding which are going to be CPU intensive and increase the costs of delivering the service. Pexip is doing that for auto zoom, but that’s because they are built as an MCU. Google decided to do this for noise suppression.
There’s packet loss concealment using machine learning now. And you can do super resolution for video to get better video quality. But in the end, all these are going to make a difference once CPUs have their own dedicated, standardized AI accelerators, like the new Apple M1 chip in them brand new Intel-less MacBooks. We just don’t have cycles to spare.
Which is why media quality has gone back to its roots. Here’s something I have in that workshop of mine:First take care of your infrastructure as much as you can to improve media quality in WebRTC
Machine learning should be added once we’re done squeezing that lemon for more performance and quality.
Google is now doing its part of optimizing the WebRTC codebase itself. It is your role to do it in your own infrastructure and application. Once done, the time will come to introduce some machine learning chops into it.
Until then? We need machine learning for two main tasks, and we see it already:
Does that mean you don’t need to invest in machine learning?
Hell no. you definitely MUST invest in machine learning.
Not for what you’ll be doing in 2021, but for what you’ll be launching in your product in early 2022. Which brings me to the heart of it all.
Machine learning is new and challenging. We’re still writing the playbook of what it means to use it for real time communications, inside a browser, using technologies such as WebAssembly.
You’ll need to decide on which use cases to invest, and what value you are going to derive of it. And you’ll need to plan for the long game here and be patient until you get results.
There’s a need to let the teams driving machine learning do the research and experimentation needed. But at the same time, they need guidance in where to look at and what to experiment with.
The post A blueprint to improving WebRTC media quality using AI appeared first on BlogGeek.me.
WebRTC growth during 2020 came in waves, just like the pandemic and its quarantines. Here how it looks and where we are all headed.
Let’s look at some interesting performance indicators of WebRTC use and adoption.
2020 is the year of video communications.
2020 is also the year of WebRTC.Table of contents
In my introductory slides of my WebRTC workshop 4 months ago, I had that as a very strong theme:
The slide above illustrates what the statistics at the time were for the big meetings vendors.
Since then, the numbers have grown. Microsoft Teams, for example, reached 115M DAU. That’s Daily Active Users.
While not all of the growth is in video calls, these services have a video focus to them.
Out of these 4 vendors:
Guest access growth for Microsoft Teams and Cisco WebEx can be attributed to some extent to WebRTC. With Google Meet, it is all WebRTC related.Gartner’s Magic Quadrant for Meeting Solutions (& WebRTC)
Gartner has its nice magic quadrant diagrams. Here’s the one just published for meeting solutions:
Which of the vendors in this magic quadrant diagram use WebRTC? I’ve marked the vendors in red for you:
The ones not marked might have WebRTC – I am just not aware of it. The ones marked have WebRTC support in production in their products. How central is it to their product is a different question though.
The thing here is that no matter what magic quadrant from Gartner you’ll be looking at for whatever market category that involves communications, WebRTC will be used as the underlying technology by many of the vendors.
Contemplating if WebRTC is the technology to use? Look at the reds above.A surge in use of WebRTC
I decided to leave the best for last.
Chrome collects and shares statistics of JS API calls in the browser and their “popularity”.
Lets look how getUserMedia() looks like:
Interestingly, we see an adoption curve where each round of quarantine raises the use of WebRTC to a higher level.
From a steady, boring 0.05% of use pre-pandemic, the new normal is settling well above 0.2% of the page loads.
How can we explain the rise from July to October? Is this a sustained growth happening as the pandemic found its second wave in different countries and social distancing gradually came back in force throughout the globe? Is it due to the fact that schools started opening around the world in August and September, many of them strictly remotely? Is it due to more services being introduced online that offer WebRTC based communications in them?
If you ask WebRTC, we’ve reached the peak of the second wave of the pandemic.Where do we go from here?
On a more serious note though, the huge surge in WebRTC traffic brought with it new use cases and a lot of learnings regarding scaling and operationalizing WebRTC.
The post WebRTC Growth – is it a back-to-school pandemic phenomena? appeared first on BlogGeek.me.
If you are planning to use WebRTC P2P mesh to power your service, don’t expect it to scale to large sessions. Here’s why.
Every once in a while someone comes in with the idea to broadcast or conduct a large scale video session with WebRTC without the use of media servers. Just using pure WebRTC P2P mesh technology.
While interesting as a research topic for university, I don’t think that taking that route to production is a viable approach. Yet.Table of contents
If you are focusing on data only WebRTC mesh, then skip to the last section of this article.
When dealing with WebRTC and indicating P2P or mesh, the focus is almost always on media transport. The signaling still flows through servers (single or distributed). For a simple 1:1 voice or video call, WebRTC P2P is an obvious choice.From a WebRTC client perspective, a 1:1 session is similar if it is done using P2P mesh or using a media server
The diagram below shows that from the perspective of the WebRTC client, there is no difference between going through a media server or going P2P – in both cases, it sends out a single media channel and receives a single media channel. In both cases, we’d expect the bitrates to be similar as well.
Making this into a group call in P2P translates into a mesh network, where every WebRTC client has a peer connection opened to all other clients directly.WebRTC mesh architecture. Or is it mess architecture? Why use WebRTC P2P mesh?
There are two main alluring reasons for vendors to want to use WebRTC P2P mesh as an architectural solution:
If WebRTC P2P mesh is so great, with cheaper operating costs and better privacy, then why not use it?
Because it brings with it a lot of challenges and headaches when it comes to bandwidth and CPU requirements. So much so that it fails miserably in many cases.
It is also important to note here that in ALL cases of 3 users or more in a call, alternative solutions that rely on media servers give better performance and user experience. Always – at least as long as the media servers infrastructure is properly deployed and configured.Bandwidth challenges in WebRTC P2P mesh
Assume we want pristine quality. Single speaker, 10 listeners.
The above layout illustrates what most users of this conference would like to see and experience. The speaker may alternate during the meeting, switching the person being displayed in the bigger frame.
As we’re all watching this on large screens (you do have a 28” 4K display – right?), we’d rather receive this at HD resolution and not QVGA. For that, we’d want at least 1.5Mbps of the speaker to be received by everyone.Strain on the uplink
In a mesh topology, the speaker needs to send the media to all the participants. Here’s what that means exactly:In WebRTC mesh, we put a bigger strain on the uplink
1.5Mbps times 10 equals 15Mbps on the uplink. Not something that most people have. Not something that I think my strained FTTH network will be able to give me whenever I need it. Especially not during the pandemic.
In an office setting, where people need to use the network in parallel, giving every user in a remote meeting 15Mbps uplink won’t be possible.
On top of that, we’ve got 10 separate peer connections to 10 different locations. WebRTC has its one internal bandwidth estimation algorithm that Google implemented in libwebrtc, which is great. But how well does it handle so many peer connections on the client’s side? Has anyone at Google ever tried to target or even optimize for this scenario? Remember – none of Google’s own services run in a mesh topology. Winning this one is going to be an uphill battle.Bandwidth estimation on the downlink
Let’s look at the viewers/subscribers/participants/users or whatever else you want to call them.
If we pick a gallery view layout, then we are going to receive 10 incoming video streams. Reduce that to 9 for layout simplicity and we get this illustration:
There are 9 other users out there who generate video streams and send them our way. These 9 streams are competing on our downlink network resources and for our machine’s attention and CPU.
Each of them is independent of the others and have little knowledge about the others.
How can the viewer understand his downlink network conditions properly? Let alone try to instruct these sends on how and what to send. A media server has the same set of problems to deal with, but it does that with two main advantages:
Then there’s the CPU to deal with in WebRTC P2P mesh.
Each video stream from our speaker to the viewers has its own dedicated video encoder. With our 10 viewers, that means 10 video encoders.
A few minor insights here if I may:
So. going for a group video call?
Want to use WebRTC P2P mesh?
You’re stuck at 300kbps or less for your outgoing video even if your network has great uplink. Because your device’s CPU is going to burn cycles on encoding multiple times.
Which also means that people aren’t going to like hearing their laptop’s fans or touch their heating smartphone (and depleting battery) on that call.Can we do better?
Probably. A single encoder would make the CPU problem a wee bit smaller. It will bring with it headaches of matching the bitrate to all viewers (each has his own network and device limitations).
Using simulcast in some manner here may help, but that’s not how it is intended to be used or how it has been implemented either.
So this approach requires someone to make the modifications to the WebRTC codebase. And for Google to adopt them. Did I already say Google has no incentive in investing in this?Alternatives to WebRTC P2P mesh
You can get a group video call to work in WebRTC P2P mesh architecture. It will mean very low bitrate and reduced video quality. But it will work. At least to some extent.
There are other models which perform better, but require media servers.WebRTC offers media server alternatives to mesh in the form of SFU and MCU
Using an MCU model, you mix all the video and audio streams in the MCU, making sure each participant receives and sends only a single stream towards the MCU.
With the SFU model, you route media around between participants while trying to balance their limitations with the media inputs the SFU receives.
You can learn more about in my WebRTC multiparty architectures article.A word about WebRTC data channel mesh
I haven’t really touched WebRTC mesh architectures for data channels.
All the reasons and challenges detailed above don’t apply there directly. CPU and bandwidth relied on the concept of needing to encode, send, receive and decode live video. In most cases, this isn’t what we’re dealing with when trying to build mesh data channel networks. There, the main concern/challenge is going to be proper creation and connection of the peer connections in WebRTC.
If what you are doing isn’t a group video call (or live video broadcast from a browser to others) then a WebRTC P2P mesh architecture might work for you. If it will or won’t is something to analyze case by case.
In the last 2 months I’ve dived into the world of CPaaS again, updating my WebRTC API focused report. Oh, and there’s a new free ebook.
There have been many changes since my last update,so this one was greatly overdue.
API platforms changed hands due to mergers and acquisitions. Vendors joining the market. Others leaving or just pivoting away from APIs.
And then we had AWS and Azure entering the CPaaS market.
What I did in these last two months was interview and review all the vendors in my report again, to see what has changed and update that part of the report. I learned a lot from the process.
As with every time where I shift focus to a certain market, I took the time to process my own thoughts by writing them down here in a series of articles.
Here are two things I wanted to share with you, as well as announce my next upcoming projects.Table of contents
I finished and published the WebRTC API report last week. The result:
Agora decided to sponsor this report (thanks a bunch!). They are one of the interesting vendors in this space, offering an IP video/voice focused platform with their own data centers spread across the globe and a lot of research done in machine learning to improve media processing.
If you are looking to learn more, then you can:
The previous 3 articles in my site here were all focused on CPaaS, looking at different angles on how CPaaS is changing.
The first one dealt with the future of CPaaS, especially considering the pandemic and how it affects everything and everyone.
In the second article, I looked at AWS Chime SDK and Azure Communication Services, trying to understand what their entry into CPaaS is going to change in the market.
For the third and last article, the focus went to Twilio Signal 2020. Considering how they redefined the market in the last 4 years in each such event, this event was a bit of a downer. It did bring with it many insights.
If you’re more into printing and reading, or sharing with others, then I packaged all of these 3 articles into one ebook, making it easier to consume.
I called the ebook CPaaS in 2020 – a market in transition. Because this is what it is…Download my CPaaS in 2020 ebook Advanced WebRTC Architecture Course – update & office hours
With my WebRTC API report now updated and finally launched, I can go back to focusing on other projects I am running.
My WebRTC Courses have been around for over 4 years now. I’ve been updating them regularly and I am doing it again to my main signature course – the Advanced WebRTC Architecture training.Updates
There are going to be 2 new lessons and around 10 lessons that are already being updated and recorded all over again.
The purpose is still to make this the best alternative out there to learning WebRTC.Office hours
Alongside the updates, I will be starting another round of office hours for the course. These will start in December.
The office hours is where students can come and learn online and in-person with me specific topics in WebRTC, as well as ask questions about anything related to WebRTC – and their own projects.
If you were thinking of learning WebRTC, then the best timing for it would be to enroll now and join the office hours. These are complementary to the course and open for anyone with a valid course subscription.WebRTC Insights – a new service
Following and catching up with everything in WebRTC is time consuming. It is also tedious. And you need to know where to look and what each bit of information means to you.
To make this a wee bit easier, I’ve decided with the help of Philipp Hancke to start a new service together – WebRTC Insights
In this service, you receive an email every two weeks. This email includes all the important changes to WebRTC
This gives you actionable insights to your own planning and reduces the risks in your development. Both Philipp and me have been doing this for a while, but doing it together brings it to a new level.
If you want to learn more and subscribe to this service, then check the new WebRTC Insights page.
Walkthrough and deep analysis of how Azure Communications Service makes use of WebRTC by Gustavo Garcia
The post How does the new Azure Communication Services implement WebRTC? (Gustavo Garcia) appeared first on webrtcHacks.
Chrome recently added the option of adding redundancy to audio streams using the RED format as defined in RFC 2198, and Fippo wrote about the process and implementation in a previous article. You should catch-up on that post, but to summarize quickly RED works by adding redundant payloads with different timestamps in the same packet. […]
Twilio Signal 2020 occurred virtually this year. The number of new announcements or market changing ones was low compared to previous years. I expected more from Twilio as the leading CPaaS vendor.Table of contents
Twilio Signal is Twilio’s yearly event where its major announcements are made. It is also a gathering place where customers, partners and even Twilio CPaaS competitors come to meet. This year, as all other events, Signal was virtual. Twilio built its own hosting platform and event experience and did a good job at that.Twilio Signal – past events
I’ve watched the keynote twice, and several of the other sessions, including all major announcement sessions. I came out of this feeling a wee bit disappointed. There was nothing really interesting or groundbreaking this year. Especially not if you compare it to some of the previous years:
In 2020, we’ve seen Twilio Microservices (the Electric Imp acquisition), Frontline, Video Go, Event Streams and Verify Push.Twilio By the Numbers
The main keynote by Jeff Lawson, Twilio CEO, had 3 components to it, with 3 main messages:
I’ll focus on the big and new parts here.
Twilio is now 12 years old and it has accomplished a lot. Jeff threw the “Twilio is big” numbers too fast for my taste, not even letting some of the big numbers register in our minds properly.
Here are the numbers. I tried aligning them with last year’s numbers from Twilio 2019:20192020Interactions750B1TUnique phone numbers2.8B3BCalls/minute32,500–Peak SMS/second13,000–Email addresses3B/quarter50%Video minutes–3BCustomers160,000200,000+Developers6M– What the numbers mean
Jeff alluded to the new normal, forced on us due to the pandemic. In many ways, this has been the main theme of Signal and the sessions.
My gripe with the “new normal” moniker to our situation is that there isn’t anything normal about it and it isn’t really here to stay.
Yes. We are seeing an accelerated move towards digital transformation and the cloud, but some of this shift, and especially the high usage in some sectors (such as education) aren’t here to stay post-pandemic.
For me, there’s no “new normal”. Just a transition to one, which will take time. How the future is going to look is hard to say from our current position.
Which leads me to the interview Jeff did with John Donahoe, Nike CEO.Nike and digital transformation
Jeff picked John Donahoe as the first person to interview during the keynote. It is an interesting choice.
I found it a tad ironic to get an explanation about social good and how Nike in all its years promoted social causes. It got me thinking about the Nike sweatshops. Other than this little history reframing that was done, the interview was quite good.
Two sentences that John said really resonated with me:
“Every business in the world is embracing digital transformation. We all have no choice”
The shift towards making businesses more digital has been inevitable.
Just think of all the on premise contact centers and what they now have to do when all of their agents are working from home. Or how all brick and mortar stores need a digital footprint to be able to even stay in business and sell throughout the quarantines.
“There is no finish line”
I should start using it myself.
There are a lot of discussions around build vs buy that I participate in, especially when it comes to the decision to build a WebRTC infrastructure versus buying an existing one via CPaaS vendors. In many cases, the argument and focus is on the initial development effort and a lot less on maintenance. The thing about maintenance is that it is almost as hard as the initial development, especially because there is no finish line – the product team will always ask for more features and capabilities which will drive more investment.Twilio Microvisor
The first announcement made during the keynote was about a new product – Twilio Microvisor.
The Twilio Microvisor is an extension of the Twilio Super SIM and its Internet of Things initiative, which many don’t even couple and view as CPaaS (I’ve been ignoring it as well).
The world of IOT and M2M is a challenging one. It includes different networks and carriers, differences in geographies and regulation, different hardware devices and chipsets.
Earlier in the year, Twilio acquired Electric Imp. This acquisition is now the Twilio Microvisor.
Up until now, the only real touching point that Twilio had with the physical world was their Super SIM. With Microvisor (and Electric Imp) that changes, and Twilio is mucking around with microcontrollers, firmware and hardware.
It the special announcements session, Evan Cummack, GM of IoT at Twilio, explained that there was a gap in the market – as a developer you either had to begin from scratch or use readymade solutions:The gap between IOT alternatives of developers: DIY or bespoke solutions
He ignored a few of the competitors for the Twilio offering, but these are less flexible and open anyways.
What Twilio is doing with Microvisor, is taking care of a few important aspects of IOT development:Twilio Microvisor features takes care of the heavy lifting of security for developers
The secure part here is key, as it is the one thing we struggle with greatly in IOT these days. This solution will remove a lot of the headaches of IOT development and get more products released.
It is also where Twilio is competing not with other CPaaS vendors but rather with cloud vendors, who also started offering IOT tooling in recent years.Twilio Video WebRTC Go
Coming from the Video and WebRTC space, this is where I am most frustrated.The need and growth of video
With the pandemic going on, Twilio had to do something about video, an area where little investment on their part has taken place. Until 2020, this has been understandable. Growth came from elsewhere and it didn’t seem like video is that important.
All this has changed. Zoom exploded, Agora.io had a great IPO, and Twilio itself saw an increase of 500% of daily usage for its video.Twilio reiterating the need and uses of video communication
The one to talk about Twilio Programmable Video was Michelle Grover, Chief Information Officer. Her part of the keynote revolved around the market need. The main market verticals here were retail and health.
It was more a reminder that Twilio is doing video than anything else.The new WebRTC announcement
The new announcement? Twilio Video WebRTC Go
What is Twilio Video WebRTC Go?
For context, pricing of 25 GB/month on Twilio’s TURN servers in the US is $10/month.
If you developed your own signaling and your own application, relying on Twilio’s TURN servers, then switching to Twilio Video WebRTC Go will save you $10.
But what you really get here is Twilio Video P2P that costs $0.0015/minute. In this configuration, you get the full infrastructure and support of Twilio’s signaling, logging and SDKs practically for free if your service is smaller than 25 GB/month of TURN media relay. How many video sessions can this accommodate? That’s something you’ll need to calculate.
For Twilio this is a win, as it gets more companies to adopt its Programmable Video at a very low price to Twilio (remember – video isn’t a serious money maker for Twilio yet, so helping these smaller users to grow their business and then have them start paying is just fine). With all the video API services out there, a free offering from a large vendor is a first. While limited, it is probably useful for many companies starting their way with 1:1 video calling.On open source and Twilio
The fact that Twilio is calling their reference apps “Open Source Video Collaboration Apps” is a bit silly. These are references/samples running on top of the Twilio Programmable Video API and are not meant, designed or easily usable on top of any other vendor or on top of any other infrastructure.
Calling a piece of code, no matter how big, open source, while forcing its user to consume other paid services in order to use it is not exactly open source.
This isn’t to say that this open source reference app isn’t useful. It surely is most useful. It gives developers a better starting point for their application, and Twilio has taken the time at Signal to offer a session titled “Accelerating Development of Collaboration Apps with Twilio Video” dedicated exactly to this.
It is a trend I see of CPaaS vendors going towards higher level abstractions. Twilio is doing that with nocode (=Twilio Studio), programmable enterprise (=Twilio Flex), reference apps for video (this one) and now with Frontline (later in this article).Nothing new under the sun here
For me this says that Twilio hasn’t invested in video as much in the last year or two. If they had, they would have announced something more thrilling and interesting. Maybe larger meetings, above 50 participants? Broadcasting capabilities? Noise suppression? Something…Twilio Flex ecosystem
The keynote and the session had a lot of Twilio Flex content in them. This is less about developers and more about contact centers.A show of force for Twilio Flex, but sharing customer logos
In this event, Tony Lama, Vice President, Contact Center Sales at Twilio mentioned in brief the fact that many features were added to Flex, but didn’t really delve into them too much. The focus was on the fact that Flex has customers and now has a thriving ecosystem of partners as well.Lots of new features, none interesting enough for the keynote
The main target for this year were the on premise contact centers – this is where Twilio is setting its sights – in the transformation these contact centers are going through as they are heading to the cloud (forced to do so earlier rather than later due to the pandemic).
This is why Twilio decided to focus on the ecosystem, making it into a big announcement:
This targets exactly the on premise contact centers, where large deployments with many agents and a lot of custom integration code and features were added over the years. An ecosystem around Flex gives Twilio the reach it needs.
It is also why Twilio introduced its latest Flex partner – Deloitte Digital – who offer system integration in this target market.
Twilio Flex and its current set of announcements is less about CPaaS and developers and more about content center as a service (CCaaS).Twilio Frontline
In that vein, the announcement of Twilio Frontline was made.
Interestingly, this was introduced by Simon Khalaf, SVP and GM, Messaging at Twilio.
Twilio Frontline is a new complete, closed, mobile application and service which enables employees in a company to directly communicate with customers through messaging channels.
The main benefits touted about Frontline? SSO (Single Sign-on) and CRM integration
This is far remote from the developer roots and target audience of Twilio, so it will be interesting to see how this plays out and redefines Twilio itself. My guess is that Frontline started as a skunk works project during the pandemic, one that turned into a new product that is now looking for a home at Twilio and within its bigger storyline.
I wonder though, was this built on top of Twilio Conversations, which was introduced at Signal 2019, or is it something implemented on top of Twilio Flex?
If this was implemented on top of Twilio Flex (which I believe it was), then why is the SVP and GM of Messaging at Twilio the one introducing it? And why wasn’t it designed, developed and even introduced as a programmable solution? Part of Flex. Maybe even an “open source application” on top of Flex.
Frontline is an interesting product. But what does it have to do with Twilio?Other announcements
There was little in the keynote of Twilio about APIs and CPaaS and more about the higher level abstractions and complete applications (Flex and Frontline). This shows a maturity level at Twilio, where most of the CPaaS domains are already well covered by their APIs.
Two additional announcements of new features/products were made, though not in the keynote itself.Twilio Event Streams
That trillion human interactions? These are probably just events in the Twilio system:
This is the slide shared in the session discussing the new feature/product of Twilio Event Streams. It isn’t a trillion but it is close enough.
What Twilio did was consolidate all of its events into a single hook, calling it Event Streams, offering a single integration point for collection of events. The first sink selected for these events is Amazon Kinesis, with more to probably be added later, based on customer demand.
Moving towards consolidated data management shows maturity and an increase in the customers that are using multiple Twilio products.Twilio Verify Push
Another new product/feature is Twilio Verify Push. This enables a mobile application to be used as a trusted device/app to validate login on another device (as well as on the device itself). The end result is reduction in the SMS volume.
While nice, I am waiting here for Google and Apple to close this gap and offer their own verification mechanisms to all instead of having application developers rely on third party services.
As for Twilio, this makes for a sensible and useful addition to their Twilio Verify service.Machine Learning was missing
What was missing at Twilio Signal 2020 is AI and machine learning.
No really interesting improvements shared about Twilio Autopilot. No cool introduction of noise suppression or other media processing machine learning capability. Nothing.
There were a few mentions on how Autopilot is used by customers during the create bots in order to deflect calls and handle the volume (nice stories that we’ve heard would be the main use case for Autopilot already).
The only “real” thing around AI? At the end of the keynote, Jeff Lawson had his short “live” coding session.Jeff, coding “live”. Still magical
This time, he went for using OpenAI’s GPT-3, a per-trained natural language processing engine. He made it understand TwiML constructs (the XML format used by Twilio sometimes) so that users can write a sentence of what they want, and the service would generate the TwiML for them. A nice toy to play with. I wonder what people would do from here with it, as it opens up a lot of questions, thoughts and ideas.
Machine learning is one of the main pillars I see in post-pandemic CPaaS offerings. Twilio has the skill set inhouse to pull this off, but they need to focus there more than they are doing today. They should probably also partner or acquire in this space to keep in pace with where the industry is headed.The coming CPaaS fight is in the enterprise
The enterprise story of Twilio came at the beginning of the keynote. Jeff wanted to make sure everyone knew and understood that Twilio is ready for the enterprise and being used by the enterprise. The careful selection of guests throughout the keynote showed that as well – they were all established enterprises. No cool startup this time. No crazy garage developers. Just formidable businesses that existed for years.Twilio is ready for the enterprise, with all the relevant certificates and procedures
I decided to leave this to the end since this is where Twilio is being challenged.
The challenge comes in the form of Amazon and Microsoft going towards CPaaS. Both of these vendors are:
Amazon will probably introduce machine learning capabilities such as noise suppression as part of its CPaaS offering soon. They have it available in Amazon Chime, so placing it in the Chime SDK is the next logical step.
Microsoft runs their CPaaS on the same infrastructure that Teams is running on. Twilio touts 3B video minutes a year while Microsoft Teams has up to 5B meeting minutes a day. I am sure that it accumulates to a considerably larger number than 3B video minutes a year.
Both Amazon and Microsoft have ways to go in stabilizing their APIs and attracting developers and attention to it. They might not be highly interested in this CPaaS business as much as Twilio is, so would probably never reach the same level of maturity and breadth of features and flexibility of Twilio. But they will surely win market share. Market share that could have easily been Twilio’s.
What is also very interesting to note is that while Amazon and Microsoft made a point of not mentioning WebRTC in the front of their CPaaS platforms (both of which are video first and use WebRTC), Twilio decided to bring WebRTC to the front with their new offering of Twilio Video WebRTC Go. I wonder which works better for enterprise sales.
Anyway, with 75% of contact centers still on premise, the enterprise market as a whole is still only starting its path towards digital transformation and with the new phrase I just adopted of “there is no finish line”, there is definitely room for growth for Twilio and its many competitors.
Interesting times ahead of our industry.
The post Twilio Signal 2020. I expected more from the leading CPaaS vendor appeared first on BlogGeek.me.
Amazon Chime SDK and Azure Communication Services mark the entrance of the cloud giants to the CPaaS space, and they are doing it from a WebRTC API angle.
Ever since Twilio became popular, a question was raised over and over again:
When will one of the large IaaS players (Amazon, Microsoft or Google) acquire them or start competing with them directly?
There was no good answer. At least not until 2020, where 3 things happened:
This. Changes. Everything.
(it doesn’t. It changes only some things, but bear with me)
I already discussed how the pandemic changes priorities for CPaaS vendors. This new development is going to make things more of a mess.Why now?
Amazon Chime SDK was already announced and launched close to the end of 2019. They already have customers and success stories under their belt. Why am I just now getting to look at how IaaS vendors are changing the market?
Probably a bit because I am doing the update to my WebRTC API platforms report this month. But also because of Microsoft’s announcement of their Azure Communication Services.Amazon Chime SDK
Amazon started the work to video communications by the introduction of Chime a few years back. Chime is an enterprise communication service (in the UCaaS space), which is akin to Zoom, Google Meet and Microsoft Teams. It enables companies to communicate internally and externally via video and voice with a better set of collaboration tools than just phone calls.
For some time now Amazon Chime was also offered as a whitelabel solution that vendors could “make their own” and integrate it with their service. But it doesn’t allow for much flexibility in terms of the workflow, business logic and user authentication. This has led Amazon to introduce the Amazon Chime SDK.
The Chime SDK is one rung lower in the stack. It enables a developer access to the logical building blocks of communications, offering a pure communication API that can be used to connect to any other service. A direct competitor to the other CPaaS vendors offering video capabilities.
What Chime SDK did to really disrupt the market was lower the price point per minute. It comes at a rate of $0.0017 per user per minute. Twilio answered with its own price drop in September 2020:A 60% reduction in Twilio Programmable Video price points
The new rates are still above the Amazon Chime SDK price points, but they are 40% their previous price points.
It should be noted that peer-to-peer calling available in Twilio Programmable Video is at $0.0015, lower than the Amazon price, but of a slightly different service and feature set.
What Amazon is “selling” here? The AWS story. From the main Chime SDK page:
AWS Lambda is already there. Connectivity to other AWS services are also part of the bigger spiel.Azure Communication Services (AKA ACS)
Microsoft just announced Azure Communication Services in a public preview. This is a full CPaaS offering that includes Video, Chat, SMS and Telephony calling. The interesting tidbits alluded to in the announcement:
Watch that video above. There’s a visual explanation of remote visual assistance. I’d never think of explaining embedded video communications or programmable video communications this way – because I am in the industry for this long. What Micsoroft is doing here is educating the market in the most basic way possible. Something we were missing in our market without even knowing it. This type of an approach can work well in the enterprise space, which hasn’t adopted such services in droves just yet.
What makes this so interesting is this:
On pricing, Microsoft was a bit more traditional and less bold than Amazon, sticking to the $0.004/minute price point the market seems to have adopted.The new model for Video CPaaS?
Even before Amazon and Microsoft joined this space, there were two objectives you could see in the mid-term and long-term roadmap for video CPaaS vendors:
These map where the new video CPaaS is headed, and the fact that Amazon and Microsoft both come with this “built-in” will accelerate things further.Machine Learning
Everyone’s doing machine learning these days, and it is part of the future of communications and WebRTC.
Amazon Chime SDK will be offering their noise suppression capabilities. Connect to Kinesis and enable access to all their other machine learning services.
Microsoft in their launch already mentioned Azure Cognitive Services as something that plays/will play nice with ACS.
Other CPaaS vendors are figuring out their way in this space as well, but part of their offering is usually how to gain access to the media for… sending it to the cloud for machine learning analysis. That cloud is going to be AWS and Azure more often than not. Being in that cloud to begin with is going to be an advantage for these cloud vendors and their CPaaS offerings.
Also remember that cloud vendors live and breath machine learning already. CPaaS vendors? Less so.Higher abstractions
Everyone in this space is talking about simplicity now.
How can I get developers to do their work in hours versus days. Days versus weeks. Weeks versus… no… weeks is too long already.
While this is unrealistic for a full fledged, polished service, it is something that works well towards an MVP or a first stab at a ready product.
Some do this by offering open source or reference applications on top of their CPaaS APIs. Others by offering this as a set of ready-made and highly configurable widgets.
It doesn’t seem like anyone has cracked the code of what is needed here, but the growing focus shows there’s something missing. Especially if we want developers to need to know less about WebRTC and media routing and more about their application logic.
I think that Amazon and Microsoft joining this market will speed up the efforts in this domain, as companies search for differentiation and quick onboarding.Why telephony is dying and communication is growing
Both Amazon and Microsoft are leading here with video, adding chat and telephony later. Later can be immediately after the initial launch, but it is still later.
In the past it made sense to do the opposite. Lead with PSTN and SMS as money makers, and add WebRTC voice and video, waiting for them to grow in adoption.
Taking the opposite approach shows where the future of consumption is.Winners
Who are the winners when CPaaS is done by the cloud vendors?Users
If cloud vendors are joining this game, it means there’s enough $$$ in this market to make it interesting, which means more users are consuming such services.
The market education that these cloud vendors are capable of doing and their reach is higher than the other CPaaS vendors, excluding maybe Twilio. This will end up with more enterprises and businesses offering such services and end users using them.Tier 1 cloud vendors
Amazon and Microsoft. Their timing couldn’t have been better.
If I haven’t known that Bill Gates is causing the pandemic so he can chip us all when his vaccine comes to market and causes all birds to fall from the sky due to 5G, I’d might end up saying that Jeff Bezos is to blame because he wanted the Chime SDK to grow in market share.
In all seriousness though, this gets both Amazon and Microsoft in front of the developers that use them for additional types of services that these developers are going to consume.Smaller cloud vendors
Digital Ocean and Oracle.
Why are they winners? I am not sure how Twilio can continue running Programmable Video on top of AWS and compete with AWS Chime SDK on price and geographic spread.
Same for the other CPaaS vendors who might be using AWS or Azure. They will be thinking hard if they want to keep their media stacks on these platforms or move them elsewhere. They can move them to Google Cloud, but Google just might introduce the same capabilities and become a competitor. Next in line will be Digital Ocean and Oracle, both cloud vendors that are carrying real time media traffic already. If I were a sales person there, I’d pick up the phone today and call the CPaaS vendors one after the other…Developers
A definite win. More choice. In clouds they already use. With a price war coming up.
What’s there to lose?Losers
Who are the losers when CPaaS is done by the cloud vendors?CPaaS vendors
They now have more competition. And not from smaller startups, but rather from the leading cloud vendors.
Cloud vendors already cater to developers, and a larger audience of developers.
Things are going to get interesting for these vendors, as they need to rethink differentiation, their own infrastructure and their pricing.Twilio
Twilio is the leading CPaaS vendor today.
They are using AWS. Everywhere.
This is definitely hurting them and will hurt them more moving forward.
Out of all the threats to Twilio, having cloud vendors competing head to head with them was the biggest one, and it is now happening.
It made sense for someone like Amazon to acquire them and use them as the communication stack for AWS. now it won’t happen.
Maybe Google will acquire them, though this seems far fetched to me.Google
3 leading cloud vendors.
Google is left behind in its communication APIs for developers, which is sad, considering they are the main driving force behind WebRTC.
I wonder if and when will Google close this gap.Developers
This will definitely rattle the existing vendors. Some of them might not make it through. So choice will again get a wee bit limited as this plays out.
While cloud vendors are great, their support isn’t the best. They tend to offer support to the smaller developers and companies through third parties and not directly, so there’s going to be less of that available. And that for a domain that is still very complex in its nature.
Developers both win and lose from this development.Updating my WebRTC API report
There’s a lot of change in the CPaaS domain. I mostly look at these vendors from a WebRTC prism, but not only.
This past month I’ve been working on updating my Choosing a WebRTC API platform report. I had a lot of briefings with the various vendors, researched their websites, added vendors, removed vendors. Grueling work.
The updated report will be published during October. It will include ~25 vendors, and touch everything from build vs buy, selection KPIs, vendor listing and pricing.
If you are looking to understand this domain better or need to select one vendor over another for an important project, then this report is for you. From today and until the report gets published, there’s a wee bit over 25% discount using coupon code API2020LAUNCH. Purchasing the report now will give you access to the current report as well as the fresh update once it is available.
The post Cloud giants joining the WebRTC API game. How is that changing the CPaaS landscape? appeared first on BlogGeek.me.
The pandemic is changing everything. CPaaS providers need to change their priorities and focus as well.
It is around this time of the year that I start thinking about where the CPaaS market is headed.
The pandemic is an epochal event. It caught the CPaaS industry somewhat ready, with gaps found in their video offerings. Behind the pandemic, a few other market changes are taking shape, affecting how CPaaS providers need to plan ahead.
I’d like to look at a few of these trends and outline what I see as the basis of CPaaS competition for the future.CPaaS features map CPaaS marketecture and features map
The diagram above shows the CPaaS features map. It is a kind of a marketecture diagram of the various bits and pieces that make up CPaaS.
I’ve layered it from Infrastructure, through Communications Building Blocks and Higher Abstraction to the Simplified Runtime domain. While not all CPaaS vendors will fill all building blocks in this map, they all see it in front of them one way or another.
Here are a few things to note:
If I had to map priorities for 2021, I’d probably create this heatmap:CPaaS areas of investment in 2020-2021 The pandemic and CPaaS vendors
In many ways, the pandemic is accelerating the need for CPaaS providers. The world switched en masse from one of physical interactions to a virtual one. This, in turn, exposed a few aspects in the CPaaS market.Digital transformation fast forward
The image above circulated on Twitter some time in March-April this year. It is spot on.
Digital transformation is here and it is here to stay. It came about a few years faster than expected and to get by, companies are relying more on communications and a lot of it comes today from vendors who use CPaaS or by developing the solutions needed on top of CPaaS platforms.
The thing is, in many cases, the increase is also catching businesses off guard, with call centers and support teams being overwhelmed with incidents. And that at a point in time where everyone is forced to work from home – including the call center agents.
This in turn, increases the requirements around technologies that assist in automation of processes and communication channels. Call deflection and agent assist solutions are taking center stage. This changes a bit how CPaaS vendors need to treat communication APIs, and especially what these APIs need to enable.
Are we looking now for more or less Uber-like solutions of matching a customer to a service provider? Or are we more about getting hold of the interaction’s content in real time and injecting insights into it, with or without a human agent?
I don’t have the answers, but I have a feeling that they are different than they were 9 months ago.CPaaS vendors totally missed video Video growth was unexpected, catching most CPaaS vendors unprepared
Yap. We had CPaaS vendors doing video. A few of them. And they’re just fine. Up until the point that video becomes important for everyone and that totally new use cases pop up in our market on almost a daily basis.
Zoom doesn’t mean a magnifying glass anymore. Nor is it talking about getting a closer look.
During the pandemic?
All of the above? Focus on video communications. None of them have any telephony roots or strong telephony capabilities. No phone numbers or SMS capabilities to speak of.
AWS decided it would be nice to join the frey, so they launched their own Chime SDK. With price points that challenge the existing players.
Twilio decided this month to lower their video price points. Cutting them down by some 60%.
8×8’s Jitsi is coming up with its own managed video API service, pricing it around MAU as opposed to the more common per minute pricing.
There’s a minor price war coming up around video APIs. It will be interesting to see how this plays out.Lack of WFH tooling in CPaaS
WFH = Work From HomeWorking from home isn’t just working from a different location
Welp… we’ve built all these nice communication services, but we’ve designed them mostly to work for the office.
On premise call centers moved to the cloud by adopting CPaaS, which is great, but the workforce itself still came to the office. All calls and communications took place from a controlled and managed environment.
The pandemic has forced call centers of tens of thousands of agents to stop coming to the office while continuing to work. From home. How do call center managers know anything about the environment of the home employee? How can he make sense of the quality of experience his agents and his customers are getting?
From the interest we see at testRTC in our qualityRTC service, there’s a real gap there.
Call this self promotion, but it is one of many areas where CPaaS vendors need to improve in order to offer a suitable WFH solution. Giving APIs is nice. Giving backend network insights and quality related dashboards is nice. Giving pre-call tests capabilities is nice. But I am not sure it is enough anymore.
Other aspects of WFH that aren’t catered for by CPaaS vendors? The need for noise suppression and background blurring/removal – to fit into the current work environments of call center agents and other workers.The pandemic will pass, but digital transformation won’t Are we really in a new normal?
It was supposed to be a quick 2 months thing. Maybe 6. A year tops.
Then came Google and Facebook (not governments, because they can’t seem to be so realistic and pessimistic with their citizens), and simply let anyone work from home at least until July 2021. At least.
Fujitsu? Decided to cut office space by 50% in 3 years as the new normal.
LivePerson, an Israeli company with 1,300 employees decided to give up on its offices altogether and go 100% WFH. This saves money and apparently most employees prefer it while management doesn’t see enough of a degradation in production output.
This obviously isn’t the case everywhere. In a recent interview with the The Wall Street Journal, Reed Hastings, CEO of Netflix had this to say about remote work:
“I don’t see any positives. Not being able to get together in person, particularly internationally, is a pure negative. I’ve been super impressed at people’s sacrifices.”
To some degree, he is correct. It greatly depends on the type of industry and company.
Dean Bubley says it best about business events:
In-person business events will rise again, although I’m less certain about office work.[…]
The #NewNormal will not be 100% remote. Once a vaccine is available, I hope that it isn’t even 50% #WFH.
My wife is a Pilates and Salsa dance teacher. She needs to work remotely now from time to time, with Zoom and recorded lessons. Her students? They’re fine with it, but whenever they can come over or do a face-to-face-in-the-flesh lesson – they’d take the opportunity.
This means that whatever it is CPaaS vendors are seeing as requirements may well stay and stick with them for the long run. What we have now isn’t a new normal, but there’s no going back to the old normal either.3 pillars of CPaaS competition and differentiation in 2021
When I had to decide what are the main areas of investment for CPaaS when it comes to differentiation and competition towards 2021, I came to these 3 domains: machine learning, video and diagnostics.
There are two reasons why I chose these domains:
Noise suppression. Background replacement. Super resolution. Bandwidth estimation. Packet loss concealment. …
All these are algorithms in the media processing domain affecting the user experience in communications. Like everything else they are now shifting towards using a lot more machine learning than in the past.
The current forerunner in importance and mindshare is noise suppression, with a lot of partnerships and M&A activities around it.
When it comes to machine learning in media quality, what are CPaaS vendors doing today?
Almost nothing at all.
The rest? Not doing much about machine learning, researching or doing bots.
This cannot last.
We’ve already seen how WebRTC is being unbundled for the purpose of differentiation. That differentiation will come in the form of optimizations, mostly done by use of machine learning.
What will vendors do? Especially when we see the leading UCaaS vendors actively investing in machine learning media processing capabilities? This sets the bar to what a communication service needs to look like, and without such capabilities, why should I as a developer use that CPaaS vendor?2# – Video, Video, Video Tony Robbins going virtual. Is this a CPaaS implementation???
Did I already say we’re in the year of the video?
A billion have been indoctrinated over a period of 1 month this year on how to use Zoom. don’t nitpick me on the exact number please. My mother now users Zoom in her daily life of a variety of activities, including a book reading club she joined
Many CPaaS vendors had video capabilities but they usually amounted to 1:1 interactions or small group sizes. There isn’t a day going by where I don’t get a new requirement from someone that CPaaS providers can’t cater for today. Many of these are in the domain of broadcasts and large groups (100 or more participants). Using CPaaS for them today feels like hacking at best. Impossibly challenging at most.
There are many areas where CPaaS providers are lacking when it comes to video. Here are the few that immediately come to mind:
The investment in video communications in all its facets will be important to stay competitive in this space.#3 – Diagnostics and analytics
It is great that you can communicate, but what happens when things go haywire?
In my recent round of updates I am doing for my Choosing a WebRTC API Platform report, many of the vendors made sure I know they have a dashboard for quality and network monitoring. Different vendors give it different names, but they all understood that unlike telephony, there’s a need for insights here, especially since networks are unmanaged.
It isn’t about me as a client understanding if the CPaaS vendor is doing a good job, but rather about me understanding my users’ networks and experience. Current dashboard solutions will need to evolve further to give the insights their customers are looking for.Didn’t you miss anything?
In my future of CPaaS article from last year I mentioned a few additional trends. Some of them have been reiterated here, though from a different angle and with a different narrative that fits better with the changing times.
There were three topics that weren’t mentioned here yet, and I want to give them a bit of room and explain where I see them in 2021 with CPaaS.nocode / low code
Still a thing. Serverless, Flow, Zapier integration, drag and drop tools. All there. All needed.
For the most part, CPaaS vendors seem to be content with the current state of affairs and the current tools they have. Investment in this domain in 2020 didn’t yield anything vastly different, new or interesting.
The domain of nocode is still relevant and interesting. For now, it seems to be mostly limited to the telephony (and voice) aspects of CPaaS.CCaaS and UCaaS
The lines are blurring elsewhere as well. Areas of IoT (below), messaging and notifications, live streaming – are all suitable adjacencies for expansion of CPaaS vendors.
The largest areas though are CCaaS and UCaaS: contact centers and unified communications
Acronyms will be tricky here. So bear with me.
In another world, just next by, other SaaS solutions are blurring their lines. Gist (the chat widget I am using on my WebRTC course site) announced to its customers that it is releasing a full fledged CRM. From conversations to CRM.
CRMs in turn, can use CPaaS vendors directly to build up their own CCaaS offering. With the higher level abstractions geared towards customer engagement, CPaaS vendors now offer a simple route for CRMs in this direction.
This will continue, though I don’t see it as direct competition or real differentiation within the CPaaS domain itself.IoT
Twilio seems to be the only CPaaS vendor investing in the Internet of Things. It acquired Electric Imp earlier this year. The acquisition wasn’t made with much fanfare, as this isn’t the main focus of Twilio and the current market is interested less in IoT than it is in video calls.
Is IoT part of CPaaS? Time will tell.
I believe that it is, but for now, only Twilio seems to be investing in that domain where none of its other immediate CPaaS competitors have the appetite for it. This will not change in the next couple of years as focus for CPaaS is elsewhere at the moment.Updating my WebRTC API report
There’s a lot of change in the CPaaS domain. I mostly look at these vendors from a WebRTC prism, but not only.
This past month I’ve been working on updating my Choosing a WebRTC API platform report. I had a lot of briefings with the various vendors, researched their websites, added vendors, removed vendors. Grueling work.
The updated report will be published during October. It will include ~25 vendors, and touch everything from build vs buy, selection KPIs, vendor listing and pricing.
If you are looking to understand this domain better or need to select one vendor over another for an important project, then this report is for you. From today and until the report gets published, there’s a wee bit over 25% discount using coupon code API2020LAUNCH. Purchasing the report now will give you access to the current report as well as the fresh update once it is available.
The post What should CPaaS providers do today to prepare for the “post pandemic”? appeared first on BlogGeek.me.
Communication vendors are waking up to the need to invest in ML/AI in media processing. The challenge will be to get ML in WebRTC.
Two years ago, I published along with Chad Hart a report called AI in RTC. In it, we’ve reviewed the various areas where machine learning is relevant when it comes to real time communications. We’ve interviewed vendors to understand what they’re doing and looked at the available research.
We mapped 4 areas:
That last area was tricky. Almost everyone was using rule engines and heuristics at the time for all of their media processing algorithms and only a few made attempts to use machine learning.
My argument was this:At some point, applying more heuristics to media processing algorithms loses its appeal
There’s so much we can do with rule engines and heuristics. Over time, machine learning will catch up and be better. We are now at that inflection point. Partially because of the technology advances, but a lot because of the pandemic.ML in media processing is challenging
When looking at machine learning in media processing, there’s one word that comes to mind: challenging
Machine learning is challenging.
Media processing is challenging.
These are two separate and far apart disciplines that need to be handled.
The data you look at is analog in its nature, and there’s often little to no labeled data sets to work with.
A few of the things you need to figure out here?
This isn’t just another checkmark to place in your roadmap’s feature list. There’s a lot of planning, management effort and research that needs to go into it. A lot more than in most other features you’ve got lined up.The noise suppression gold rush
If I had to pick areas where machine learning is finding a home in communications, it will be two main areas:
Both topics were always there, but took centerstage during the pandemic. People started working from weird places (like home with kids) and you now can’t blame them. One of the best games I play in workshops now? Checking who’s got the most interesting room behind him…
Video background is about stopping me from playing. Noise suppression is about you not hearing the lawn mower buzzing 16 floors below me, or the all-too-active neighbor above me who likes to home renovate whenever I am on a call – with a power drill.How I think my neighbor looks like whenever I am on a conference call
This need has led to a few quick wins all around. The interesting 3 taking place in the domain of WebRTC (or near enough) are probably the stories about Google Meet, Discord/Krisp and Cisco/BabbleLabs.Google Meet Google Meet built its own noise suppression technology
In June, Serge Lachapelle, G Suite Director of Product Management was “called to the flag” and was asked to do a quick interview for The Verge on Google Meet’s noise suppression. Serge was once the product manager for WebRTC at Google and moved on to Google Meet a few years back.
You can watch the short interview here:
The gist of it?
As I stated earlier, Google isn’t taking any prisoners here and contributing this back to the community freely as part of WebRTC. They are making sure to differentiate by making sure their machine learning chops are implemented outside of the open source WebRTC library. This is exactly what I’d do in their place.Discord Discord “bought” its way to noise suppression by partnering with Krisp
Krisp is one of the few vendors tackling machine learning in media processing and doing that as a product/service and not a feature. They’ve been at it for a couple of years now, and things seems to be going in their favor this year.
Krisp managed to do a few things:
The Discord story was first published in April on Discord’s blog. Noise suppression was added in beta to the Discord desktop app. Was that done using the browser technology used in Discord’s Electron app or by the native implementation that Krisp has is an open question, but not the most relevant one.
Three months later, in July, Discord got noise suppression into iOS and Android. This was also done using Krisp and with a spanking short video explainer:Ongoing success
Here are my thoughts here:
As we continue to improve voice chat, Krisp is an integral part of making Discord your place to talk. No matter how stressful the world around us may be, Krisp is here to help every one of our 100 million monthly active users feel more connected to our far-away friends.
My read of it? Krisp might be acquired and gobbled up by Discord to make sure this technology stays off the hands of others – if that hasn’t happened already – just look at this page – https://krisp.ai/discord/ (and then compare it to their homepage).Cisco Cisco gobbled up BabbleLabs to own noise suppression technology
In the case of Cisco, the traditional approach of reducing risk by acquiring the technology was selected – acquihiring.
Last week, Cisco issued a press release of their intent to acquire BabbleLabs.
BabbleLabs was in the same space as Krisp. A company offering machine learning-based algorithms to process voice. The main algorithm there today as we’ve seen is noise suppression. This is what Cisco were looking for and now they will have it inhouse and directly integrated into WebEx.
Cisco devices not to self-develop. They also decided to own the technology. The reasons?
Will BabbleLabs stay open? No.
In his recent post about the acquisition, Chris Rowen, CEO of BabbleLabs, explains what lead to the acquisition and paints a colorful future. The only thing missing in that post is what about existing customers. The answer is going to be a simple one: They will be supported until the next renewal date, when they will simply be let go.
A win to Krisp. If it isn’t in the process of being acquired itself already.Who’s next?
This definitely isn’t the end of it. We will see more vendors taking notice to this one and adding noise suppression. This will happen either through self-development or through the licensing of third party solutions such as Krisp.
The challenge with these third party solutions is that they feel more like a feature than a product or a full fledged service. On one hand, everyone needs them now. On the other hand, they need to be embedded deep in the technology stack of the vendors using them. The end result is relatively small companies with a low ceiling to their potential growth (=not billion dollar companies). This puts a strain on such companies, especially if they are VC backed.
On the other hand, everyone needs noise suppression now. Where do they go to buy it? How do they build it?Noise suppression is just the beginning
Noise suppression is just the beginning here. In the workshop I did last month on WebRTC innovation and differentiation, I’ve taken the time to focus on this. How machine learning is now finding a place in bringing differentiation to the actual communication. Noise suppression was one of the topics discussed, with many others.
There were 3 main areas that we will see growing investment in:
Each of these domains has its own set of headaches and nuances.Server, native or browser? Should you employ ML in WebRTC in the cloud or on the edge?
This is a big question.
If you look at the examples I’ve given for noise suppression:
Going for native or browser means you’re closer to the edge and the user. You can do things faster, more efficiently and with a lower cost to you (you’re practically employing the user’s device to bear the brunt of running the machine learning inference algorithm). That also means you have less resources for other things like the actual video and you’re limited in the size of the model you can use for your algorithm.
Cloud means a central place where you can do training, inference, A/B testing, etc. It is probably easier to maintain and operate in the longer run, but it will add some delay to the media and will definitely cost you to run at scale.
Each company will choose differently here, and you may see a company choosing for one algorithm to run it in the cloud and for another to run on the edge.Are you planning for this ML/AI future?
Machine learning and artificial intelligence is in our future. Both in the communication space and elsewhere. It is finally coming also to media processing directly. In a few years, it will be a common requirement from services.
Are you planning for that in any way?
Do you know how you’re going to get there?
Will you be relying on third parties or on your own inhouse technology for it?
There are no open source solutions at the moment for any of it. At least not in a way that can be productized in a short timeframe.
If you need assistance with answering these questions, then check my workshop. It is recorded and available online and it is more relevant than ever.
Back in April 2020 a Citizenlab reported on Zoom’s rather weak encryption and stated that Zoom uses the SILK codec for audio. Sadly, the article did not contain the raw data to validate that and let me look at it further. Thankfully Natalie Silvanovich from Googles Project Zero helped me out using the Frida tracing […]
2020 marks the point of WebRTC unbundling. It seems like the new initiatives are the beginning of the end of WebRTC as we know it as we enter the era of differentiation.
Life is interesting with WebRTC. One moment, it is the only way to get real time media towards a web browser. And the next, there are other alternatives. Though no one is quite announcing them the way they should.
We’re at the cusp of getting WebRTC 1.0 officially released. Seriously this time. For real. I think. Well… maybe.Towards differentiation
If I were to chart our path through this crazy world of WebRTC, it would look something like this:2020 marks the beginning of the differentiation stage for WebRTC
Towards the end of 2019, and at greater force during the pandemic, we’ve seen how the future of WebRTC looks like. It is all about differentiation.
Up until now, all vendors had access to the same WebRTC stack, as it is implemented by Google (and the other browser vendors), with the exact same capabilities in the browser.
I’ve alluded to it in my article about Google’s private WebRTC roadmap. Since then, many additional signals came from Google marking this as the way forward.
Today, there are 2 separate WebRTC stacks – the one available to all, and the one used internally by Google in native applications. While this is something everyone can do, Google is now leveraging this option to its fullest.
The interesting thing that is happening is taking place somewhat elsewhere though. WebRTC is now being unbundled so that Google (and others) don’t need to maintain two separate versions, but rather can have their own “differentiation” implemented on top of “WebRTC”.Unbundling WebRTC
At this point, you’re probably asking yourselves what does that mean exactly. Before we continue, I suggest you watch the last 15 minutes from web.dev LIVE Day Two:
That’s where Google is showing off the progress made in Chrome and what the future holds.
The whole framing of this session feels “off”. Google here is contemplating how they can bring a solution that can fit Zoom, that when 99% of all vendors have figured out already how to be in the browser – by using WebRTC.
The solution here is to unbundle WebRTC into 3 separate components:The components set to unbundle WebRTC
While these can all be used for new and exciting use cases (think Google Stadia, with a simpler implementation), they can also be used to implement something akin to what WebRTC does (without the peer-to-peer capability).
WebTransport replaces SRTP. WebCodecs does the encoding/decoding. WebAssembly does all the differentiation and some of the heavy lifting left (things like bandwidth estimation). Echo cancellation and other audio algorithms… not sure where they end up with – maybe inside WebCodecs or implemented using WebAssembly.What comes after the unbundling of WebRTC?
This isn’t just a thought. It is an active effort taking place at Google and in standardization bodies. The discussion today is about enabling new use cases, but the more important discussion is about what that means to the future of WebRTC.
As we unbundle WebRTC, do we need it anymore?
With Google, as they have switched gears towards differentiation already, it is not that hard to see how they shift away from WebRTC in their own applications:Google Stadia Does Stadia has a reason to use WebRTC?
Google Stadia is all about cloud gaming. WebRTC is currently used there because it was the closest and only solution Google had for low latency live streaming towards a web browser.
What does Google Stadia need from WebRTC?
That’s a small portion of what WebRTC can do, and using it as the monolith that it is is probably hurting Google’s ability to optimize the performance further.
Sending back user actions were already implemented in Stadia on top of QUIC and not SCTP. That’s because Google has greater control over QUIC’s implementation than it does over SCTP. They are probably already using an early implementation of WebTransport, which is built on top of QUIC in Stadia.
The decoding part? Easier to just do over WebTransport as well and be done with it instead of messing around with the intricacies of setting up WebRTC peer connections and maintaining them.
For Stadia, unbundling WebRTC will result moving away from WebRTC to a WebTransport+WebCodecs combo is the natural choice.Google Duo & Google Meet Meet & Duo. Will moving away from WebRTC improve their competitiveness?
For Duo and Meet things are a bit less apparent.
They are built on top of WebRTC and use it to its fullest. Both have been optimized during this pandemic to squeeze every ounce of potential out of what WebRTC can do.
But is it enough?Differentiation in WebRTC
Google has been adding layers of differentiation and features on top and inside of WebRTC recently to fit their requirements as the pandemic hit. Suddenly, video became important enough and Zoom’s IPO and its huge rise in popularity made sure that management attention inside Google shifted towards these two products.
This caused an acceleration of the roadmap and the introduction of new features – most of them to catch up and close the gap with Zoom’s capabilities.
These features ranged from simple performance optimizations, through beefing up security (Google Duo doing E2EE now), towards machine learning stuff:
Can Google innovate and move faster if they used the unbundled variant? Instead of using WebRTC, just make use of WebTransport+WebCodecs+WebAssembly?
What advantages would they derive out of such a move?
If I were Google, I’d be planning ahead to migrate away from WebRTC to this newer alternative in the next 3-5 years. It won’t happen in a day, but it certainly makes sense to do.Can/should Google maintain two versions of WebRTC?
Today, for all intent and purpose, Google maintains two separate versions of WebRTC.
The first is the one we all know and love. It is the version found in webrtc.org and the one that is compiled into Chrome.
The other one is the one Google uses and promotes, where it invests in differentiation through the use of machine learning. This is where their WaveNetEQ can be found.
Do you think Google will be putting engineers to improve the packet loss concealment algorithm in the WebRTC code in Chrome or would it put these engineers to improve its WaveNetEQ packet loss concealment algorithm? Which one would further its goal, assuming they don’t have the manpower to do both? (and they don’t)
I can easily see a mid-term future where Google invests a lot less in WebRTC inside Chrome and shifts focus towards WebTransport+WebCodecs with their own proprietary media engine implementation on top of it powered by WebAssembly.
Will that happen tomorrow? No.
Should you be concerned and even prepare for such an outcome? That depends, but for many the answer should be Yes.The end of a level playing field and back to survival of the fittest
WebRTC brought us to an interested world. It leveled the playing field for anyone to adopt and use real-time voice and video communication technologies with a relatively small investment. It got us as far as where we are today, but it might not take us any further.Recent changes marks the shift from a level playing field in WebRTC towards survival of the fittest
For this to be sustainable, browser vendors need to further invest in the quality of their WebRTC implementations and make that investment open for general use. Here’s the problem:Apple (Safari)
Doesn’t really invest in anything of consequence in WebRTC.
Apple cares more about FaceTime than all of that WebRTC nonsense anyways…Mozilla (Firefox)
Actually have a decent implementation.
Their latest Edge release is Chromium based.
And then there’s Microsoft Teams, which offers a sub par experience in the browser than it does in the native application. All of the investment of Teams is going towards improving quality and user experience in the app. The web is just an afterthought at the momentGoogle (Chrome)
Believe WebRTC is good enough.
Now that we’re heading towards differentiation, the larger vendors are going to invest in gutting WebRTC and improving it while keeping that effort to themselves.
No more level playing field.Prepare for the future of WebRTC
What I’ve outlined above is a potential future of WebRTC. One that is becoming more and more possible in my mind.
There’s a lot you can do today to take WebRTC and optimize around it. Making your application more scalable. Offering better media quality as well as use experience. Growing call sizes to hundreds or participants or more.
Investing in these areas is going to be important moving forward.
I’ve recently created a workshop covering the present and future of WebRTC, along with techniques and best practices employed by vendors in this space. If you want to learn more, you may want to take that workshop.
The post WebRTC unbundling: the beginning of the end for WebRTC? appeared first on BlogGeek.me.
WebRTC IP addresses and port ranges can be a bit tricky for those unfamiliar enough with VoIP. I’d like to shed some light about this topic.
A recent back and forth discussion that I had with one of the people taking my online WebRTC course made it clear to me that there are still things I take for granted because I come with a VoIP heritage to what it is I am doing today. Which is why this article here.
Connecting a WebRTC session takes multiple network connections and messages taking place over different types of transport protocols. There are two reasons why that decision was made for WebRTC:
Lets see how connections get made over the internet and how WebRTC makes use of that.A quick explainer to internet connections
We will start by looking at the building blocks of digital communications – TCP and UDP.
The table below summarizes a bit the differences between the two:TCP and UDP are two extremes of how transport protocols can be expressed TCP connections
TCP is a reliable transport protocol. As such, it has a built-in retransmission mechanisms that is meant to make sure whatever is sent is received on the other end and in the same order of sending.
To do that properly, a TCP connection needs to be created. A TCP connection is a set of 4 values:
Source IP:Source port + Destination IP:Destination Port
How does one establish a TCP connection?
On your local machine you “bind” one of the local IP addresses of the machine to a local port number. That IP and port needs to be available and not taken for something else already. Then you need to try and connect to the destination IP:port.
Let’s say I want to connect to google.com.
For me, google.com resolves to the IP address 22.214.171.124. Assuming I want to connect to port 80 (a “randomly” picked port), I’d do the following: Bind a local IP:port (arbitrary local port), and connect it to 126.96.36.199:80.
Knowing the IP:port on source and destination of the connection means knowing the connection – there cannot be two such connections. Once you bind a local port to connect it to a remote address over TCP, that port cannot be reused until the connection is closed and done with.
If I want to open another TCP connection from my machine to the same address, I will need to bind yet another port on my local address and connect it to the destination IP:port,
Obviously, there are some caveats and edge cases I am ignoring here, but for our needs, the above is enough of an explanation.UDP “connections”
Since UDP is connectionless, there’s no real connection with UDP. No context whatsoever.
To send a message over UDP, I need again the quad of values:
Source IP:Source port + Destination IP:Destination Port
But this time, there’s no real connection. What happens here is that I open a local IP:port, and whenever I want to send out a message, I just tell it the destination IP:port and be done with it.WebRTC signaling connections and addresses
WebRTC signaling is just like any other web application connection.
In order to send and receive the SDP blobs to make the connection, I need to be able to communicate between the browsers and that is done using traditional networking means available in the browser: either HTTP or WebSocket. Both (ignoring HTTP/3) are implemented on top of TCP.
What does that mean?
The end result?The signaling server has a static IP and port, while the client is “dynamic” in nature
Local ports are arbitrary (and ephemeral). Destination port is 443 (or whatever advertised by the server).WebRTC media connections and addresses
Media in WebRTC gets connected via SRTP. Most of the time, that would happen over UDP, which is what we will focus on in this section.
In naive SRTP implementations from before the WebRTC era, each video call usually used 4 separate connections:
What happens though is this –
Since these addresses and ports are local, there’s high probability that they will be blocked by firewalls for incoming traffic.
Media servers work in the exact same way. In most cases, the addresses that they will use will be public IP addresses, but the ports will be arbitrary. That’s because media servers usually prefer handling each incoming device separately, by receiving its traffic on a dedicated socket connected to a specific port.STUN “connections” and addresses
Since we’re all behind NATs with our private IP addresses, we need to know our public IP address so we can connect to others directly (peer-to-peer).
To do that, STUN is used. WebRTC will take the media local IP:port it created (in that section above), and use it to “connect” over UDP to a STUN server.
This is in concept somewhat similar to how our signaling works – the local IP address has an arbitrary port, while the remote IP:port is known – and configured in advance in our peer connection iceServers. My advice? Have that port be 443.
Why do we do all that with STUN? So that we create a pinhole through the NAT which will allocate for us a public IP address (and port). The STUN server will respond back with the IP address and port it saw, and we will publish that so that the other side will attempt reaching out to us on that public IP:port pair. If the NAT allows such binding, then we will have our session established.The STUN server has a static IP and port, while the client (and NAT) operate with “dynamic” IP addresses and ports
The above shows how Google’s STUN server works from my machine in AppRTC:
With TURN, the server is relaying our media towards the other user. For that to happen, my browser needed to:
The above shows how Google’s TURN server works from my machine in AppRTC:
UDP, TCP and TLS work similarly in TURN when it comes to address and port allocation. What is important to notice here is how the TURN server opens up and allocates ports on the its public IP address whenever someone tries to connect through it.Understanding port ranges in WebRTC configurations
WebRTC makes use of a range of addresses, ports and transport protocols. Far more than anything else that we run in our browsers. As such, it can be quite complex to grasp. There is order and logic in this chaos – this isn’t something inflicted on you because someone wanted to be mean.
In WebRTC the addresses and ports that get allocated by the end devices (=browsers), media servers and TURN servers are dynamic. This means that in many cases we have to deal with port ranges.
Go to any voice or video conferencing service running over the Internet. Search for their address and port configuration. They all have that information in their knowledgebase. A list of addresses and ports you need to open in your firewalls, written nicely on a page so that the IT guy will be able to copy it to his firewall rules.
Should these ranges be large? As in 49,152 to 65,535? Should this range be squeezed down maybe?
I’ve seen vendors creating a port range of 10 or 100 ports. That’s usually too little to run in scale when the time comes. I’d go with a range of 10,000 ports or more. I’d probably also try first to estimate the capacity of the machine in question and figure out if more ports might be needed to maintain the sessions per second I am planning on supporting (allocated TCP ports take some time to clear up).
Is this “wholesale” port range a real security threat or just an imaginary one? How do you go about explaining the need to customers who like their networks all clamped down and closed?
If you are looking to learn more about WebRTC, check out my WebRTC training courses. In the near future, I will start working on a new course about TURN installation and configuration – if you are interested in early access – do let me know.
The post WebRTC ports: Understanding IP addresses and port ranges in WebRTC appeared first on BlogGeek.me.
WebRTC TURN servers are an essential piece of almost any WebRTC deployment. If you aren’t using them, then make sure you have a VERY good reason.
Connecting a WebRTC session is an orchestrated effort done with the assistance of multiple WebRTC servers. The NAT traversal servers in WebRTC are in charge of making sure the media gets properly connected. These servers are STUN and TURN.3 ways to connect WebRTC sessions
When connecting a session between two browsers (peer-to-peer) in WebRTC, there are 3 different alternatives that might happen.Connect directly, across the local network Connecting WebRTC over a local network
If both devices are on the local network, then there’s no special effort needed to be done to get them connected to each other. If one device has the local IP address of the other device, then they can communicate with each other directly.
Most of the time and for most use cases, this is NOT going to be the case.Connect directly, over the internet, with public IP addresses Connecting WebRTC directly using public IP address obtained via STUN
When the devices aren’t inside the same local network, then the way to reach each other can only be done through public IP addresses. Since our devices don’t know their public IP addresses, they need to ask for it first.
This is where STUN comes in. It enables the devices to ask a STUN server “what is my public IP address?”
Assuming all is well, and there are no other blocking factors, then the public IP address is enough to get the devices to connect to each other. Common lore indicates that around 80% of all connections can be resolved by either using the local IP address or by use of STUN and public IP addresses.Route the media through a WebRTC TURN server Connecting WebRTC by using TURN to relay the media
Knowing the public IP address is great, but it might not be enough.
There are multiple reasons for this, one of them being that the NAT and firewall devices in use are not allowing such direct traffic to take place. In such cases, we route the data through an intermediary public server called TURN.
Since we are routing the data, it is an expensive endeavor compared to the other approaches – it has bandwidth costs associated with it and it is why you Google won’t ever offer a free TURN server.Transport protocols and WebRTC TURN servers
TURN comes in 3 different flavors in WebRTC (6 if you want to be more accurate).How testRTC checks and explains connectivity alternatives of TURN servers in qualityRTC
You can relay your WebRTC data over TURN by going either over IPv4 or IPv6, where IPv4 is the more popular choice.
Then there’s the choice of connecting over UDP, TCP or TLS.
UDP would work best here because WebRTC knows best when and how to manage network congestion and if to use retransmissions. Since it doesn’t always work, it might require the use of TCP or even TLS.
Which type of a connection would you end up with? You won’t really know until the connection gets established, so you’ll need to have all your options opened.When is a TURN server needed in WebRTC?
That’s easy. Whenever there can’t be a direct connection between the two devices.
For peer to peer, you will need to install and run a TURN server.Try direct, then TURN/UDP, then TURN/TCP and finally TURN/TLS
The illustration above shows our “priorities” in how we’d like a session to connect in a peer to peer scenario.
If you are connecting your devices to a media server (be it an SFU for group calling or any other type of a server), you’ll still need a TURN server.
Why? Because some firewalls block certain types of traffic. Many just block UDP. Some may even block TCP.
With a typical WebRTC media server, my suggestion is to configure TURN/TCP and TURN/TLS transports and remove the TURN/UDP option – since you have direct access to the public IP address of the media server, there’s no point in using TURN/UDP.Try direct to server, then TURN/TCP and finally TURN/TLS
The illustration above shows our “priorities” in how we’d like a session to connect with a media server.What about ICE-TCP?
There’s a mechanism called ICE-TCP that can be used in WebRTC. In essence, it enables a media server to provide in the SDP a ICE candidate using a TCP transport. This means the media server will actively wait on a TCP port for an incoming connection from the device.
It used to be a Chrome feature, but now it is available in all web browsers that support WebRTC.
This makes the use of TURN/TCP unnecessary, but will still leave us with the need of TURN/TLS.Try direct UDP to server, then direct ICE-TCP to server and finally TURN/TLS
The illustration above shows our “priorities” in how we’d like a session to connect with ICE-TCP turned on.The elusive (mis)configuration of TURN servers in WebRTC
Configuring TURN servers in WebRTC isn’t an easy task. The reason isn’t that this is rocket science. It is more due to the fact that checking a configuration to ensure it works properly isn’t that simple.
We are used to testing things locally. Right?
Here’s the challenge – in WebRTC, trying it on your machine, or with your machine and the one next to it – will ALWAYS WORK. Why? Because they connect directly, across the local network. This means TURN isn’t even necessary or used in such a case. So you never test that path in your code/configuration.
What can you do about it?
The above things can be done locally and repeatedly, so start there. Once you get this to work, move towards the internet to check it there.Quick facts Do you need a TURN server if you connect your sessions to a WebRTC media server?
Yes. WebRTC media servers don’t support TLS type of transport. Sometimes they do support TCP via ICE-TCP. For the cases where calls can’t connect in other ways, you will need to use TURN/TCP or TURN/TLS.Do media servers need to have WebRTC TURN server configuration?
Usually not. In most cases you will be installing media servers with direct internet access on a public IP address. This means that having TURN configured only on the WebRTC client side is enough.How do you test a TURN server configuration for your application?
An easy way is to block UDP traffic and see if your WebRTC client can still connect. Another one is to use Google’s Trickle ICE sample.
The post WebRTC TURN: Why you NEED it and when you DON’T need it appeared first on BlogGeek.me.
Once a month, I will be publishing along with Philipp Hancke a WebRTC fiddle of the month as a free lesson in my WebRTC Codelab. This continues an old tradition of Mozilla’s Jan-Ivar, the fiddle ot the week.
Time for a new experiment. If all goes well, we will be making it a monthly thing.
Somehow, a week or two ago, we came to the conclusion that it would be nice to do a short video explainer of something that people are trying to figure out with WebRTC.
To make this happen, we decide together what to do the short lesson about, then Philipp Hancke writes a jsFiddle piece of code to implement it. And then we sit and record the explanation of it, creating a new free lesson in our joint WebRTC Codelab course.What does each WebRTC fiddle of the month include?
Each WebRTC fiddle of the month has these 3 resources:
Creating the WebRTC Codelab was fun. We’re thinking of recording a new course at some point, but until we wrap our heads around that one, we’ve decided to continue recording some more lessons.
Seems like we work good together, so finding yet another excuse to do something made enough sense.
Oh, and if you happen to decide that you need to learn more about WebRTC, and end up enrolling to our course, then that’s a definite winTwo requests I have for you #1 – Check our first “fiddle of the month”
Our first explainer fiddle is about creating a peer connection that includes screen sharing and an audio microphone stream wrapped nicely together. You might get zoom-fatigue (or the WebRTC equivalent of that term), but you almost always want to talk when trying to screen share and collaborate.
Go watch our fiddle – Sharing screen + microphone together
Future WebRTC fiddles won’t be announced here, but only on social media and in the WebRTC Weekly (so subscribe to it). These fiddles will all be available in the WebRTC fiddle of the month section of the WebRTC Codelab course.#2 – Suggest some more ideas for such fiddles
Got ideas or requests for fiddles?
I can’t promise we will record your request, but I can promise you we will seriously think about it once we sit down to decide what to record.
I wanted to add local recording to my own Jitsi Meet instance. The feature wasn’t built in the way I wanted, so I set out on a hack to build something simple. That lead me down the road to discovering that: getDisplayMedia for screen capture has many quirks, mediaRecorder for media recording has some of its […]
The post Using getDisplayMedia for local recording with audio on Jitsi appeared first on webrtcHacks.
2020 offers an interesting viewpoint to WebRTC browser support. Where exactly is it available in desktop and mobile, and what can you do about it as a developer?
This is almost a yearly article that I now write, each time with a slightly different focus to it. We’re now halfway into 2020, and things are changing fast.
Here’s a quote that I am seeing a lot this year:
Sometimes this quote is quite literally true. pic.twitter.com/SkVooF9Fez— The Long Now Foundation (@longnow) January 1, 2020
It rings true for the last few weeks when it comes to WebRTC, but somehow, in the domain of WebRTC browser support, we’re still standing in place.
My most up to date slide on WebRTC browser support?
We will get back to it in detail a bit later.
For now I’d like to look at the “Can I use” website, filtered for WebRTC. It gives a good starting point (although somewhat misleading). I will use that as the basis of looking at WebRTC on desktop and mobile.WebRTC support on desktop
On the desktop today, all modern web browsers support WebRTC.
This has been the case for quite some time now. I’ve announced that this means that WebRTC is ready towards the end of 2018.
Because the consumption model in the desktop today is done through web applications, while on mobile, it is predominantly based on native applications. So the moment all desktop browsers are nicely represented and supported, things look bright.
This isn’t to say that there aren’t challenges with WebRTC browser support – obviously there are.
I can list a few of them here out of the top of my head:
When it comes to mobile, support for WebRTC is a bit more complicated.WebRTC iOS Safari support Is WebRTC really available on iOS Safari?
iOS Safari has been supporting WebRTC since Safari 11.
We’re now in Safari 13.5 and things are still rather grim when it comes to true support of WebRTC.
iOS Safari WebRTC is such a broken mess that my going suggestion to clients unfortunately is to not support it and redirect users to a native app installation. I had to manually go through all open WebRTC bugs in webkit to figure out how to explain this to my clients and help them in reaching that conclusion and even conveying that to their customers.
There are nasty bugs in iOS Safari that have been opened since 2019 or earlier relating to media handling of WebRTC. These aren’t just edge cases, but rather things you’ll have users bump into in regular use. Some of them have finally been fixed in the latest 13.5.5 beta earlier this month.
Oh – and if you plan on using any OTHER browser on iOS then WebRTC won’t be supported there. Why? Because Apple hasn’t made WebRTC available in its Webkit Webview on iOS and they aren’t allowing anyone to build a mobile iOS browser that doesn’t use Webkit as its rendering engine. So much for freedom and choice.
Up until now, there was no serious way to run a WebRTC web application in iOS Safari in production at scale. Hopefully, this is now mostly solved…Android browsers support for WebRTC How well is WebRTC supported in the fragmented Android world?
Android has its own set of headaches when it comes to WebRTC. That’s because there’s no single Android out there, but rather a slew of them.
Here’s what we can glean from a close look at that “can I use” list above.
While WebRTC is nicely supported in Android, it is going to be hard sometimes to decide what that support exactly means. Knowing that is a mix of understanding the device and the browser the web application is being executed on top.Where do we go from here?
If you read everything until here, then understand this: WebRTC is a work in progress.
It is the best (and only) alternative you have for real time communications that works in the browser without any installation. It works well enough for large companies to release applications (web and native) that attract massive user bases.
As with many other technologies, starting to use it is simple. Getting it to a professional level requires a lot more investment and commitment.
Next week I’ll be starting off my “future of WebRTC” workshop. This workshop is going to cover many aspects of the changing landscape of WebRTC. I’ll be touching issues of infrastructure, optimization and differentiation. All with a view of the current best practices as well as the latest trends.
There are 2-3 more seats available in the workshop. If you are interested in joining, check here.
VP9 is the best unused codec today that can improve video quality and media experience in your WebRTC application. Lets see who is this codec good for.
Last year there were 3 video codecs available in browsers for WebRTC: VP8, H.264 and VP9. Now there seem to be 5, with the addition of HEVC and AV1. Let’s put some sense into what is really going on, and where does each of these fit, focusing on VP9.WebRTC video codec support by browser All modern web browsers today support WebRTC
In the good old days, WebRTC video codec support was “simple”. The industry was bickering and fighting between VP8 or H.264 until resolving the matter by mandating both VP8 and H.264 codec support in web browsers.
Then Google went ahead, adding VP9 into the mix. Mozilla went along with it and added it to Firefox.
After that, we’ve got to the point where the Alliance of Open Media was created with AV1 as its video codec, prepping us nicely into the next codec war we’re going to face – the upcoming AV1 codec versus HEVC – both of which are now available (or eminently available, or soon to be available) in web browsers with WebRTC.
I’ve created this simple table for you to understand in which web browser which video codec is available for WebRTC:
To make things simple, if you need to launch something in 2020, then the video codecs available to you are:
Which leads us to the next question…HEVC & AV1. Should you join the experiment? Be sure you want to be part of WebRTC’s future video codec(s) experiment(s)
Let’s check the new video codecs that are sprouting in WebRTC to understand where we stand with them.HEVC
It seems that Apple is adding HEVC to Safari. It is available to some extent in the Safari Technology Preview, so developers can tinker with it without knowing when it will be publicly available in Safari. And there is no indication or an inkling of an indication that other browser vendors are going to join – Google won’t. Mozilla definitely won’t. Microsoft just might, but that would mean forking away a bit from Chromium which is now the engine inside their Edge browser – not the right focus in my mind for Microsoft here.
HEVC is like VP9 in the same way that H.264 is like VP8:
The only difference is that HEVC doesn’t exist in any browser yet and will only be available on Safari while H.264 is available in all browsers.
To understand where we’re at with H.264 vs VP8 we only need to read the stats shared by Google in their recent semi-celebration for 10 years of WebM and WebRTC:
“These technologies have succeeded together, as today over 90% of encoded WebRTC video in Chrome uses VP8 or VP9.”
The bolded marking is my own doing – and just so we’re clear:
Now with AV1 coming up and the huge backing behind it, the HEVC track is all but dead. At least for the majority of WebRTC developers.
If you want to use HEVC in WebRTC, then you limit yourself to future Safari releases and native applications (where you modify the WebRTC codebase to add HEVC). Don’t expect it to work in any other web browserAV1
AV1 is the best next thing in video coding. The best invention since sliced bread. The best unlikely cooperation amongst industry co-opetitors moving away from royalty bearing video codecs towards an open video codec.
It is supposed to be better than both VP9 and HEVC from a compression standpoint.
And it is supposed to be a coombaya experience where everyone is supporting it. The members list of the Alliance of Open Media foundation behind it is impressive. It includes all browser vendors and many chipset vendors. How can you go wrong here?
The only problem for me is adoption time. It takes a long time to get a video codec to market.CodecYear startedAgeH.264200317VP8200812HEVC20137VP920137AV120191Video codec maturity
To get a video codec to market properly time is needed. From specification, to implementation, to modifying the implementation to work for real time communications, to optimizing the implementation to work reasonably on available CPUs.
In Chrome it doesn’t officially exist. It is there behind a flag, making sure users can’t really enjoy it and web developers don’t have meaningful enough access to it.
Getting a codec that came out of the oven a year or two ago to production is risky business.
If you need this article to learn about codecs, then AV1 should NOT be in your roadmap in 2020. You better wait this one out a little bitWho is using VP9 codec today?
Not Google it. Google.
They use VP9 in Google Meet. That’s a large traffic source using VP9, but it says a lot about the adoption of VP9 so far.
There are also a few instances where VP9 is used in streaming or live streaming use cases. Nothing major though.Adoption challenges of using VP9 codec Photo by Mathias Jensen on Unsplash
Why so low an adoption after being out on the market for 7 years?
I can only guess…
I’ve written about the role of VP9 in WebRTC before.
The premise of VP9 is improving encoding compression over VP8.Compression rate
That comes at a cost of expending more CPU, giving us the option to balance between using network and CPU resources.VP9 gives you either less bitrate for the same quality or more quality for the same bitrate than VP8
When looking at the higher end of the bitrate equation, one may prefer using VP8 or H.264 – we have enough network resources so we couldn’t really care on that front while saving on CPU might be beneficial
On the lower end of the bitrate equation, we’d want to squeeze every bit we have running on the network on higher quality. And then using VP9 might make more sense: since the bitrate is limited, we can spare more CPU on that and use VP9.
Sad thing is we can’t really know this in advance, at least not always.Scalability
Implementing a workable large scale video group call with WebRTC isn’t trivial. There are a lot of aspects to deal with both from a network perspective as well as from a CPU perspective.
The name of the game in this case is optimization, and this comes by having more flexibility in the tools you can use for optimizing the hell out of your video experience.
The flexibility here comes from VP9 SVC implementation in WebRTC:
If I had to chart the flexibility of a WebRTC video codec for large group sessions based on the tools it gives developers, this is what I’ll get:
We expend more CPU on VP9 but we win in network performance and scalability of a video group call by doing that.Plotting a route towards AV1
AV1 is the future.
Should you skip VP9 and just head to AV1 once it is ready? I don’t know.
The thing is, we had 2020. A pandemic that got us all cooped up at home doing video calls like crazy. The world has changed and with it priorities in our industry.
We’ve fast forwarded roadmaps by 5 years, so the future is already here. Can you wait a year or two more before you introduce a better video codec? If yes, then go straight to AV1. If you can’t, then you should seriously consider starting off with VP9 adoption.A quick recap Who is using VP9 codec in WebRTC applications?
Google Meet makes use of VP9 codec. Sadly, there is no other popular, large scale WebRTC application that makes use of it.Does VP9 codec support SVC in WebRTC?
Yes. In fact, VP9 is the only codec today that supports SVC (Scalable Video Coding) in WebRTC. This gives developers more flexibility in large group video calls and even live broadcasts than other video codecs.
The challenge is that VP9 SVC support in WebRTC isn’t official or well documented.
Better compression rate compared to VP8 and H.264 which are mandatory to implement in WebRTC.
Better scalability. It has flexible tools that assist in scaling video group calls.
HEVC and AV1 don’t yet exist in WebRTC browser implementations. At least not in a way you can utilize in a production service.
VP9 is available and usable in Chrome, Firefox and Edge.
If you are looking to improve video quality or reduce bitrates in your WebRTC application, then you should seriously look at VP9.
The post VP9 Codec: Is it time to adopt it in your WebRTC application? appeared first on BlogGeek.me.
My journey creating a scalable SBC as a Service for Microsoft Teams Direct Routing is over at Snapsonic.com.
Enabling large group video calls in WebRTC is possible, but requires effort. WebRTC CPU consumption requires optimizations and that means making use of a lot of different techniques.
Cramming more users in a single WebRTC call is something I’ve been addressing here for quite some time.
The pandemic around us gave rise to the use and adoption of video conferencing everywhere. Even if this does slow down eventually, we’ve fast forwarded a few years at the very least in how people are going to use this technology.It all started with a Gallery View
What’s interesting to see is how requirements and feature sets have changed throughout the years when it comes to video conferencing. When I first joined RADVISION, the leading screen layout of a video conference was something like this:
It had multiple names at the time, though today we refer to it mainly as gallery view (because, well… that’s how Zoom calls it).
Somehow, everyone was razor focused on this. Cramming as many people as possible into a single screen. Some of it is because we didn’t know better as an industry at the time. The rest is because video conferencing was a thing done between meeting rooms with large displays.
It also fit rather well with the centralized nature of the MCU, who ruled video conferencing.Enter Speaker View
At some point in time, we all shifted towards the speaker view (name again, courtesy of Zoom):
There were a few vendors who implemented this, but I think Google Hangouts made it popular. It was the only layout they had available (up until last month), and it was well suited for the SFU technology they used. It also made a lot of sense since Hangouts took place in laptops and desktops and not inside meeting rooms.
With an SFU, we reduce the CPU load of the server by offloading that work to the user devices. At the same time, we increase our demand from the devices.Hello 2020
Then 2020 happened.
With it came social distancing, quarantines and boredom. For companies this meant that there was a need for townhall meetings for an office, done on a regular interval, just to keep employees engaged. Larger meetings between room meetings because even larger still as everyone started joining them from home.
The context of many calls went from trying to get things done to get connected and providing the shared goals and values that are easier to achieve within the office space. This in turn, got us back to the gallery view.
Oh… and it also made 20+ user meetings a lot more common.3 reasons why WebRTC is a CPU hog
The starting point is challenging with WebRTC CPU use when it comes to video calling. WebRTC has 3 things going against it at the get go already:#1 – Video takes up a lot of pixels
1080p@30fps is challenging. And 720p isn’t a walk in the park either.
The amount of pixels to process to encode 1080p?
62 million pixels every second… I can’t count that fast
You need to encode and decode all that, and if you have multiple users in the same call, the number of pixels is going to grow – at least if you’re naive in your solution’s implementation.
What does this all boils down to? WebRTC CPU use will go over the roof, especially as more users are added into that group video call of yours.#2 – Hardware acceleration isn’t always available
Without hardware acceleration, WebRTC CPU use will be high. Hardware acceleration will alleviate the pain somewhat.
Deciding to use H.264?
Going with VP8?
What about VP9?
So all these pixels? Software needs to handle them in many (or all) cases.#3 – It is general purpose
WebRTC is general purpose. It is a set of APIs in HTML that browsers implement.
These browsers have no clue about your use case, so they are not optimizing for it. They optimize for the greater good of humanity (and for Google’s own use cases when it comes to Chrome).
Implementing a large scale video conference scenario can be done in a lot of different ways. The architecture you pick will greatly affect quality but also the approach you’ll need to take towards optimization. This selection isn’t something that a browser is aware of or even the infrastructure you decide to use if we go with a CPaaS vendor.
And it all boils down to this simple graph:Complexity vs group size in WebRTC conference calls
The bigger the meeting size, the more WebRTC CPU becomes a challenge and the harder you need to work to optimize your implementation for it.Why now?
I’ve been helping out a client last month. He said something interesting –
“We probably had this issue and users complained. Now we have 100x the users, so we hear their complaints a lot more”
We are using video more than ever. It isn’t a “nice to have” kind of a thing – it is the main dish. And as such, we are finding out that WebRTC CPU (which people always complained about) is becoming a real issue. Especially in larger meetings.
Even Google are investing more effort in it than they used to:April 17, 2020 3 areas to focus on to improve performance in group video conferences
Here are 3 areas you should invest time in to reduce the CPU use of your group video application:#1 – Layout vs simulcast
Simulcast is great – if used correctly.
It allows the SFU to send different levels of quality to different users in a conference.
How is that decision made?
Think about it. And see where you can shave off on the bitrates. The lower the total bitrate a device needs to deal with when it has to encode or decode – the lower the CPU use will be.#2 – Not everyone’s talking
Large conferences have certain dynamics. Not everyone is going to speak his mind. A few will be dominant, some will voice an opinion here and there and the rest will listen in.
Can you mute the participants not speaking? Is there an elegant way for you to do it in your application without sacrificing the user experience?
There are different ways to handle this. Anything from a dominant speaker, through the use of DTX towards automatic muting and unmuting of certain users.#3 – UI implementation
How you implemented your UI will affect performance.
The way you use CSS, HTML and your JS code will eat up the CPU without even dealing with audio or video processing.
Look at things like the events you process. Try not to run too much logic that ends up changing UI elements every 100 milliseconds of time or less on each media track – you’re going to have a lot of these taking place.My eBook on the topic is now available
If you are interested in how to further optimize for video conference sizes, then I just published an eBook about it called Optimizing Group Calling in WebRTC. It includes a lot more details and suggestions on the above 3 areas of focus as well as a bunch of other optimization techniques that I am sure you’ll find very useful.
The post Surviving WebRTC CPU requirements in large group calls appeared first on BlogGeek.me.