Juniper's Koley Embraces the Edge

Bikash Koley came to Juniper from Google, but had previously worked at Ciena, so he wasn't unfamiliar with the telecom space by any means. So he likely knew what he was getting into at a manufacturer whose router and security revenues were under fire on all sides, from traditional rivals such as Cisco and from newer players not burdened by a hardware legacy.

In part one of this Q&A, Koley stressed Juniper's credentials in the software space and its commitment to open source. In part two, he argues that Juniper is better prepared than its competitors to thrive at the edge of the network, where he sees multiple distinct use cases and says Juniper and its Contrail platform are ready for both. And Koley makes the case that Juniper's security strategy is also future-proofed by enabling distributed security for the virtualized world. (See Juniper CTO: Open Source Software Can Be Profitable.)

Carol Wilson: Juniper is part of the ONF [Open Networking Foundation] initiative, as you announced in June, that is focused on setting open source reference designs for what happens at the edge of the network. How is the rise of edge computing impacting Juniper and is it clear to you where the "edge" actually is and what it will look like going forward?

Bikash Koley: For us, the use cases for edge computing are pretty clear. You can broadly break it down to two distinctive use cases and sometimes I see them getting mixed up. One is where you are virtualizing your edge network infrastructure and you're operating that on an edge cloud because it needs to be close to the edge. This is going to be fundamental to how 5G gets rolled out. The reason is primarily economics. Because if you look at the economics of 5G, it would be very hard to monetize all that has been spent to turn up a 5G network if you are not able to just build as is needed, and edge cloud gives you the ability to build as is needed.

But then there's a second use case, which is how do I allow people to run applications that need to be closer to the users and by doing so, I monetize my ability to have space and power close to the user. The perfect examples of second use cases would be IoT [and] connected cars, where latency matters and you need to be close to the users. These two are somewhat distinct use cases and from Juniper's perspective -- we are actually very active in both.

CW: Can you explain what you are doing?

BK: The first use case is going to be primarily about mobility. If you go through Contrail's history, Contrail is probably the most deployed telco NFVi/SDN software out there with 200-plus customers, 40-plus are Tier 1/Tier 2 service providers all across the globe, and most of the actual production NFVi services, a vast majority of them use Contrail as NFV/SDN.

5G is a natural extension for us and we have been preparing for that for a long time. How are we preparing for it? If you look at the common characteristics of edge cloud, there are a few things that you need. You need a small footprint, because we don't actually have a lot of space and power. We have been optimizing Contrail to fit into basically a single appliance for the whole stack, which it does today. You can't run control plane locally because again it takes too much space and power and compute. Contrail has been supporting a centralized control plane, including when the control plan runs on public cloud, for quite some time.

You need deep implementation of security because you're going to have IPsec and SSL encryption that you have to terminate, which Contrail does natively, as well as we have virtualized most of our firewall and routing devices to actually do that. On the edge, you have no option but to run microservices because full-blown services are too heavyweight to completely run on the edge. They have to be micro-service based. Contrail today has one of the best [microservice] implementations, so you can spin up microservice on containers. We do that today already for edge as a matter of fact. We have some of the first containerized VNFs [virtual network functions] that are orchestrated on Contrail. When it comes to service provider infrastructure adopting virtualization on the edge, we are very well prepared. We actually believe we are probably the most prepared when that deployment happens and we've already seen some deployments of that.

CW: Where do you see service providers deploying, other than 5G?

BK: Service providers have a distinct advantage of having an abundance of space and power that are really close, within milliseconds, of users. That is where they're differentiated from many of their competition, including public cloud. But for that to work, what they really need is, again, a cloud stack that they can deploy on the edge, that offers modularity and has all the same behaviors that I described before. Small footprint, embedded security, able to run control plane remotely, and the ability to orchestrate multiple users.

There, we have what we call Contrail distributed multicloud. The whole approach has been how do we take the micro-segmentation and multi-tenancy that we have built for core applications and how do we extend that to edge? We already have, including having the ability to securely connect your edge tapestry of core centers so that if you're doing microservices, you have a seamless infrastructure all the way from edge to core.

We also still have use cases for that already for IoT and the connected car environment. Also in terms of standardization, there are a lot of good works that are already starting to happen in Linux Foundation. There are some that are happening as part of ETSI and IEEE, so we're following them very closely. You're going to see us joining several of those as well, as they mature and we understand which ones are going to go and have the most ecosystem and market traction.

— Carol Wilson, Editor-at-Large, Light Reading

This is an edited version of a story that was originally published on Telco Transformation's sister site, Light Reading. To see the full story, click here.