Serverless Opinions

Before the Serverless Panel session, I asked the panelists to submit a 5 minute video on their “Serverless Opinions.” This was to provide a level-set for the panelists, and enable us to get into the meat of the discussion as quickly as possible.

It’s a great resource, and in this weblog I’ve summarized their opinions. The Serverless Opinions video playlist is here.

Sebastian Schumann, Network Production Development, Technology & Innovation at Deutsche Telekom

Summary: Learning about the role serverless can play, cautiously migrating towards cloud / containerization. By the time they’re ready for serverless perhaps a new approach to telephony could be possible – event driven edge compute?

Sets out a simple taxonomy of the core RTC functions (media/signalling), programmable comms (CPaaS, HTTP layer), and above that web services (that can naturally fit in a serverless paradigm). For example an EPG guide, it’s event driven and highly variable on load.

Covers the layers of abstraction and the learning his organization is going through. The key term you used is ‘control’. Serverless means you have to give up some control.

Historically the focus has been hardware-up. So for planning (financial), architecting scalability, availability, and performance everything is under your control. This mindset has limited even the move to containers, never mind public cloud. And the shift will take years, and technology will likely have moved on. Even programmatic provisioning is a significant step, and has challenges wrt financial planning

Rebuilding Telecoms: What does the customer want? What if we created an event-driven decentralized stack?

Simon Woodhead, CEO Simwood

Summary: Makes sense for some workloads, particularly VAS, as part of a hybrid architecture across on-net, cloud, and serverless.

Cloud is just somebody else’s computer. The thing that has improved is tooling (e.g. Kubernetes for orchestration), which enables on-net to run hotter and burst into the cloud (hybrid cloud).

Serverless is different to cloud, it delivers on the promise of cloud. Great example on the database query result being returned in 20 seconds that had taken weeks on-net. As Uber is to cloud, serverless is to teleportation – see above 20 sec versus 2 weeks example.

The Future is hybrid: on-net, cloud, and serverless. Right tool for the right job, integrated so hybrid works transparently.
Telephony has always been serverless.

Cloud / serverless has a high BS factor  these days because of marketing / chancers. However, compute intensive VAS are a natural fit: transcription, sentiment analysis, AI-assisted communications.

Tim Panton, Co-founder and CTO at Pipe

Summary: Lifetime of serverless functions limits its role in RTC. But taking a broader definition, serverless could be a way to rethink communications, by using the processing power at the edge.

Serverless is like a perfume company using DHL for logistics.

With serverless: define a function (the ones they have available) and let AWS or Google do the rest, no infrastructure management.
But the lifecycle of serverless functions is short, max 9 mins. Which is not compatible with the longer sessions of communications.

Unconvinced serverless will work for RTC because of its session based nature.

Tim thinks a definition of serverless beyond AWS/Google using edge devices for functions like transcoding, noise reduction, background replacement, face detection, transcription. Similar to Sebastian’s comment on Rebuilding Telecoms. What if we created an event-driven decentralized stack focused on customer needs?

João Camarate Silva, Founder & CTO GoContact

Summary: To be relevant to developers, serverless is necessary for tech companies. CPaaS is a small step in the disruption of telecoms disruption. HTTP/3 RIPP is heralding a broader shift from SIP/RTP to UDP and event based. Joao made 2 versions, a short, shown immediately below; and a longer version shown at the end of this weblog.

Serverless characteristics: don’t manage infrastructure, upload code, scales automatically, as does price.
Challenges: Control, Flexibility, Vendor lock-in, cost uncertainty.

CPaaS is a half-way house to serverless. So serverless is not new. And as Tim points out, Tropo got their first 😉

However, containerization migration is underway (cloud tooling), with telecom lagging as it’s still quite VM / HW centric.

What is driving change in the industry are web developers, thanks to the HTTP layer of CPaaS (where growth and value are being created).

Web developers are focused on fashion / minimize effort to deliver code. Hence Docker not IT / DevOps. And we’re increasingly seeing serverless winning web developer attention. Google / AWS / Azure spend heavily on dev marketing.

As Tech Companies to attract developers we need to be relevant to their world, and serverless is part of it. This applies to hiring of developers as well.

CPaaS is just the first step in the disruption of telecoms, which is ripe for change.

Open source HTTP/3 RIPP work could herald a fundamental shift, currently focused on interconnect. Broader move from SIP/RTP to UDP/RIPP and event based across all telecoms?