Planned Outages? Where we’re going, we don’t need planned outages
Recently the Nokia IP Routing team announced the Nokia 7730 SXR, a new family of service and aggregation routers for CSPs and mission-critical network builders that’s powered by a new silicon processor, the Nokia FPcx and running the SR Linux Network Operating System (NOS).
In this follow-up blog, which borrows its title from a line from "Back to the Future", we'll focus on one of the key benefits that the new FPcx and SR Linux bring to IP routing, in-service upgrades.
I’ve been in and around IP and routers since the early ‘90s having started my networking career as a support engineer. This meant installing and configuring customer routers and performing many NOS upgrades. In the 90s that involved physical engineering with being onsite at midnight and physically removing the router from the rack and the cards from the chassis. It also meant a set of manual tasks with the programing of PROMs and the replacement of the operating systems via chip pull and insertion. A scary task when you have one new set of PROMs (144 of them in one upgrade I did) and under an hour to upgrade the router.
Over time these upgrades became easier, technologies like flash memory and redundant control planes meant that upgrades could be done remotely with minor impacts on the customer traffic. But these upgrades came at a cost.
To perform hitless upgrades required redundant control planes in the router chassis, that increased the CapEx of the routers and increased the demands on site with more power and rack space required. It also didn’t always create hitless upgrades.
With redundant control planes, control plane traffic could be protected and that ensured that IP Application layer protocols like BGP or OSPF etc. would remain stable, but the upgrade could still require a change to the data plane of the router, via new microcode deployed into the physical interfaces or line cards of the router.
To upgrade when a router requires a microcode enhancement in the data plane requires dual homing of the customer traffic to two routers, and a micro-outage to force the traffic off its primary path to the other router. Aside from the complexity of both configuration and network operations this also drives up the costs with more power, more rack space, and a doubling of the network interfaces.
With the 7730 SXR the Nokia IP team wanted to address this problem head on; create a hitless upgrade on a non-redundant control plane router.
The solution involves two key developments. Firstly, a new NPU architecture that supports independent processor clusters whereby portions of the data plane capacity could be taken offline, new microcode loaded and then put back into service, all without losing a single customer packet. This is of course only possible with a 100% programmable data path enabled via a network processor architecture; something to which Nokia has had unwavering commitment.
Secondly, that any higher layer IP protocols, like the BGP or OSPF etc. could be modularized within the NOS and independently upgraded without having to signal to neighboring routers that they were in maintenance mode or were withdrawing routes.
The key to this functionality comes from the network operating system (SR Linux) and the network processor unit (FPcx) working in concert, something that truly only happens when the same manufacturer creates both silicon and software.
The 7730 SXR silicon and software teams worked together in the initial design phases of the new router family to solve this long-standing network issue, and it took their equal parts to deliver this innovation.
The results are an industry first for service routing that delivers significant operational and capital investment benefits to CSPs and mission-critical network builders. The immediate benefits are with smaller, more fit for purpose routing platforms at the access and aggregation edge layers of the network for a start. But the bigger benefits will come with the ongoing lifecycle savings in both plant (power and rack space) and operational simplicity.
Its early days for the technology and the Nokia IP team has a strong plan on adding operational tools to the silicon/software foundation of the 7730 SXR to deliver ‘hitless’ to our customers. Ultimately the team wants to remove the need for planned outages wherever the 7730 SXR is deployed, and in a world where there is an increasing need for always on IP services it’s a challenge that Nokia IP has firmly in their sights.