There’s no suitable cloud host for WebRTC. Yet.
Earlier this year, I tried giving 2 different perspectives about WebRTC hosting on AWS. One positive and the other negative.
Since then I’ve been asked once in awhile where should people host their services. I’ve also looked at how the WebRTC API Platform vendors host their own data centers (this is part of the report on the topic).
This weekend, going over the blogs I follow, I came across this interesting post by David Mytton on GigaOm. It is about data centers and location:
It’s clear that location is going to be a competitive advantage, one where Microsoft currently holds first place but will lose it to Softlayer soon. Given the amount of money being invested, it will be interesting to see where cloud availability expands to next.
Go read that post. It is a good one. You’ll notice some interesting (and worrying) information there.
The reasons David gives for the need for more data center locations are 3: redundancy, data protection and user latency.
In the case of WebRTC, user latency is critical – not necessarily for the signaling, but if server side media processing is required, or just TURN servers to route the media, then the location of the TURN server is really important.
The interesting thing is that from the table David gives on the locations of data centers from the various cloud players, there are regions that don’t have enough data centers or have nothing at all. And none of the big players have good answers for a lot of locations (I know – people ask me). Amazon with AWS, isn’t even leading the pack – they seem to be lagging behind.
Why is this important?
We seem to think at times that the game for WebRTC may be over in some domains. Be it video conferencing services or API platforms (there are many of these already), but the thing is that for most, the game hasn’t even begun.
WebRTC is sensitive to latency, and as such requires a good spread in terms of data centers. No single cloud vendor today can be used to get that spread, leaving WebRTC players to either focus on specific geographies for their target market, or work on agnostic solutions that can span multiple cloud providers.
It also means that the room is still wide open for players who wish to provide a truly global and widespread accessibility for real time services around the globe.
Areas that are challenging to most include South America, Africa, China and countries in Asia Pacific.
Would someone please take the challenge and go figure out a solution for us all?
I come from a N. American and a SIP background, so we have always heard from SBC and softswitch vendors to centralize the infrastructure. One large Canadian telco was planning to have 2 Data centres, one in east and the other in the west to cover the entire residential voice service. Seems to me they were still thinking client-server or mainframe, when we should be thinking of distributed systems.
Back to cloud providers, perhaps SoftLayer because of their bare metal servers. Signalling can run on virtualized systems but for now media processing shouldn’t be running on hypervisors if you intend to scale.