Salsify – A New Architecture for Real-time Internet Video

Hi all -- Salsify co-author here. Surprised to see us here again, but happy to be part of the conversation (here&#x27;s a previous one: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16964112" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16964112</a>).<p>This work was led by my student Sadjad Fouladi. If you liked Salsify, you might really like Sadjad&#x27;s talk in June at USENIX ATC about &quot;gg&quot;, his system for letting people use AWS Lambda as a rented supercomputer (e.g. he can compile inkscape and Chromium really fast by outsourcing the computation to 8,000 Lambda nodes that all talk to each other directly over UDP): <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=Cc_MVldSijA" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=Cc_MVldSijA</a> (code here: <a href="https:&#x2F;&#x2F;github.com&#x2F;StanfordSNR&#x2F;gg" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;StanfordSNR&#x2F;gg</a>)<p>You might also be interested in our current video project, led by my student Francis Yan, on trying to improve live video streaming. If you visit and watch some live TV you can help contribute to our study: <a href="https:&#x2F;&#x2F;puffer.stanford.edu" rel="nofollow">https:&#x2F;&#x2F;puffer.stanford.edu</a>

From what I understand, Salsify requires a unique connection for each device being served a stream. Therefore (like WebRTC) you can&#x27;t cache frames in a CDN; and each viewer requires a connection to the origin. This gets a little expensive (which is one of the reasons why WebRTC isn&#x27;t very common relative to HTTP streaming). Have you explored any kind of multicast or cache-friendly variation of Salsify?

Earlier:<p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16964112" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16964112</a><p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16802079" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=16802079</a>

If this somehow fixes the situations where I have to click pause and un-pause to get a video to actually play, that’d be a step forward for humanity...

REMB (actually the correct name would be GCC) is the de-facto rate control and bandwith congestion estimation algorithm that&#x27;s used by WebRTC (I have some short notes with links to the algorithms that were proposed, in [0]), and it&#x27;s a technology that for all practical effects has been kind of abandoned since 2014.<p>So it&#x27;s only natural that it should be possible to greatly improve upon what we&#x27;re using today. I&#x27;m glad to see advances in this field, because it&#x27;s sad that video calls are still a hugely worse experience compared with good old phone calls...<p>[0]: <a href="https:&#x2F;&#x2F;doc-kurento.readthedocs.io&#x2F;en&#x2F;latest&#x2F;knowledge&#x2F;congestion_rmcat.html" rel="nofollow">https:&#x2F;&#x2F;doc-kurento.readthedocs.io&#x2F;en&#x2F;latest&#x2F;knowledge&#x2F;conge...</a>

Is anyone working on putting these ideas into a working product?<p>I&#x27;m taking guitar lessons over Skype and for the most part it works just fine. However, every once in a while we have to resort to recording and sending snippets over the wire to get the required fidelity. With digital amp simulators this is thankfully a trivial exercise, but it would be great to do without.<p>I&#x27;ve looked for other alternatives, but couldn&#x27;t really find any that fit the bill. The last thing I looked at was a few Audio over IP products, but all of them were designed to be run on a LAN.

I&#x27;m surprised that sites that serve videos--like YouTube and Netflix--haven&#x27;t already come up with this idea, of using a codec optimized for transmission over the network.<p>I wonder if we can do this with WebSockets? (The decoder might need to be in WASM for performance to be worthwhile.)

A little bit of latency comes from bufferbloat where network device buffers cause latency to increase.<p><a href="https:&#x2F;&#x2F;www.bufferbloat.net" rel="nofollow">https:&#x2F;&#x2F;www.bufferbloat.net</a>

Glad to see Sprout-related work going on Keith + Sadjad! I remember being amazed by sprout a few years ago.

Anyone have suggestions what protocol should be used for in-browser streaming cloud gaming?

So, how is this better or worse than RTSP and&#x2F;or h.265?

TL;DR: they created a VP8 implementation that is purely functional and exploited that property to integrate control loops of codec and transport protocol. Sounds promising.

is this worse or better than zoom?

Middle out