Adventures in Real-Time Communications in the Cloud, Manuel Pombo, GoContact.
This is an excellent introduction to the core issues that lead carriers to have a significant legacy burden, and why change is slow. In Sebastian Schumann’s Serverless opinion, he also covers these points and more.
Manuel then gives a masterclass on moving RTC workloads into the cloud. They’ve experienced VMs competing for resources with other customers’ workloads. Which has driven them to bare-metal instances and the use of SR-IOV (Single-Root Input/Output Virtualization).
Once in the cloud ecosystem, there are loads of managed services available; they highlight PostgreSQL (relational database), redis (in-memory data structure store), and Gluster (scalable network filesystem). And have found to achieve the necessary level of performance can raise costs 2-3 times.
On the media server side with bare metal they’re able to bind a public IP to a VM, which is not possible in the cloud. However, in FreeSWITCH it can autodetect the IP in use for the VM, which can get around the issue. There’s loads of great insights Manuel shares in getting RTC working in the cloud: security, load balancing, disaster recovery, the hybrid approach they’ve taken, and a peek into RIPP (Realtime Internet Peering Protocol). It’s a masterclass 🙂
I would also like to highlight a white paper written byJoão Camarate Silva, CTO and founder of GoContact, that covers the issues discussed and more, Serverless Telecommunications. As João said in his keynote, GoContact is a communication service provider with its feet firmly on the ground in telco, but its head in the clouds 🙂
Slideshare is having difficulties today, here’s a direct link to Manuel’s slides.
Description from Agenda
Manuel Pombo will share an insightful review of real-time communications in the cloud. Serverless will be a theme through TADSummit. And we’re fortunate to get this broad review to kick off our discussions.
- Highly Available and Scalable Real Time Communication (RTC) Workloads in the Cloud
- Floating IP Pattern between active-standby stateful services
- Load balancing for scalability and high availability with WebRTC and SIP
- Cross-regional DNS-based load balancing and failover
- Persistent Storage for data durability and high availability
- Dynamic / auto scaling with event-driven serverless computing
- HA WebRTC video
- HA SIP Trunking
- Best Practices
- SIP Overlay
- Monitoring
- DNS for load balancing and floating IP for failover
- Multiple Availability Zones
- Traffic management and workload placement
- Optimizing for networking
- Security, security, security
- Pros and cons of serverless
- A view of the future of RTC in the Cloud
Manuel, thank you for an excellent review of your RTC in the cloud adventure. What has been the biggest surprise through your RTC in the cloud adventure?
Thank you Manuel, great talk! I have a question regarding public clouds. In your presentation, all 3 bigger public cloud players are present in each step, that means that you have integrated your solution in all of them, or at least some kind of PoC? If yes, which of them you would recommend?
Sounds reasonable, that Hybrid besides problably require more effort, is the way to go, where most of the traffic is handled on premise and just for given clients or needs the public cloud (the 2 or 3 times more expensive figure say very clear…).
Hi Jesus. We have integrated with AWS (for being the innovative leader) and Azure (because it’s the new On-Premises where most enterprises are now migrating their datacenters too), but Google Cloud is on it’s way too.
You can’t go wrong with any of these 3, as their offer is very similar in terms of price and feature coverage.
Hi Alan. I guess our biggest surprise during this process is seeing that high quality RTC is now perfectly achievable on top of public cloud providers, as they have now matured their offer in all necessary ways.