CSPs are in a critical period of transformation right now. Omnipresent connectivity is driving new disruptive business models which are further driving up demands on the networks. Virtualization of appliances and systems is seen as a necessary step to add the agility to meet these increasing and evolving service demands. As vendors and CSPs are faced with building these virtualized systems, it’s imperative to look at the software engineering methodologies that the IT industry has successfully applied to challenges at comparable scale.
Microservices architecture has emerged over the last few years as a way to address these large scale engineering challenges. Already adopted by large organizations like Google, Amazon and Netflix, microservices provide an agile and cost-effective way for CSPs to build their services quicker and keep them running longer.
To better explore this topic, Volt Active Data teamed up with VanillaPlus to host a webinar titled “Microservices are Everywhere: What CSPs can Learn from IT”. We have reproduced some of the questions and answers below. For those that missed the webinar, you can view it here in its entirety.
Questions were voiced by George Malim of VanillaPlus, and answered by Volt Active Data’s Seeta Somagani and Analysys Mason’s John Abraham, with some comments from George. Their answers are prefaced with their name.
These answers have been edited for clarity and grammar.
Q1: Telcos in general love standards. Is there an initiative to define a consensus on what “cloud native” means when evaluating virtual network functions?
[John Abraham] I don’t think so. There is not a consensus yet. Operators are pushing a few NFV agendas to the architect, there are solutions from the cloud. A recent research report we did showed that a few operators are pushing for the development of their own cloud-native VNF. Proximus and Swisscom, to name a few, put out a call for NFV startups, to encourage cloud-native VNF development. It’s accepted that new tooling is required, but there’s no consensus yet on standards.
Q2: How can you architect microservices and avoid increased latency due to multiple services replacing a monolithic service?
[Seeta Somagani] That’s a good question. Instead of in-procedure calls to other modules, you are breaking services up into HTTP calls, which causes extra latency. This is where a highly-transactional database helps. All the data can be put in a central location, which helps correlate your requests, and creates a consistent location to look up your data. Instead of each service having its own datastore that does its own thing, using a central store helps in reducing latency.
Q3: What do you think is the timeline for mainstream adoption of cloud-native technologies?
[John] I’ll answer that question from an operator’s perspective. I think mainstream adoption is at least 5 years out. That may seem like a long time, but with believers on one side, skeptics on the other, and hybrid users somewhere in the middle, mainstream adoption will come when the hybrid users jump on the bandwagon. I suspect it will take 4-5 years for us to get there.
[George Malim] The question here I guess is how you describe mainstream, and a very large chunk of that hybrid group is the tipping point.
[Seeta] That’s a good answer John, and I agree with George about the mainstream question. I think the maturity of the products available to Telcos is important, and I believe that there is one mature product out there. I would want to see more solutions in this area to say this is mainstream and the market is maturing. For timeline, I agree with John: 5 years.
[George] I want to dial in a bit closer, and instead of “mainstream”, when do you think more products will enter the market? Mid next year, or even as early as first quarter?
[Seeta] Nokia is ahead of the curve here, and we’re going into new markets in Asia, where new CSPs are looking to go into NFV. Nokia already has a cloud-native approach. With this customer perspective, I think next year cloud-native products will gain momentum.