The future of routers requires no planned outages

This Industry Viewpoint was authored by Vinai Sirkay, Vice President, IP Routing at Nokia

If, like me, you have been in the telecommunications field for some time, you will remember that networks were traditionally architected to deliver to a service level. We used complex calculations to derive measurements such as 99.999 percent or “five nines” availability, which became the basis for contracted service-level agreements (SLAs).

Business customers knew the throughput and availability they needed and paid accordingly, especially for services that provided connectivity for their operations.

Lately, there has been a disconnect (no pun intended) between network availability and the always-on digital world we live in. It stems from the use of the internet as the digital pathway for our business, personal communication and entertainment needs.

With our accelerating consumption of digital services, there is increasing reliance on internet access as the transport link to these cloud-based services. This puts availability stresses on the access and aggregation networks that communication service providers (CSPs) built for what was originally designed to be a best-effort service.

As residential, commercial or government consumers of these cloud-based services, we expect to be able to access them 24/7 regardless of where we are—at home, on the train or in the office. So, when our connection to Microsoft Teams, Netflix, our internet banking app or Fortnite servers is broken, our world stops.

The value has shifted for us as consumers. We value the services we consume from the cloud and forget that the underlying connectivity to these services isn’t 100 percent guaranteed.   

Now don’t get me wrong: The core infrastructure that connects us to the internet, and the backbones that link the global internet to the cloud are very reliable (despite the plethora of bad actors and their DDoS bots). But at some point in the connection, most likely in the first- and second-mile networks, architectural, physical plant and economic conditions mean there are non-redundant elements. These networks were built to deliver to a service level, but we now expect them to be available at all times.

This disconnect between the consumed service and the connection, coupled with the always-on world we live in, creates a problem for CSPs. How do they update the IP network infrastructure that’s connecting all their mobile, residential and business customers to the cloud?

The easy answer is to do it during scheduled network outage windows: “Service will be affected between the hours of 2 a.m. and 6 a.m. on Sunday.”

This may have been the default answer in the past. But that quiet period on the network doesn’t exist anymore. On any given Sunday morning, there could be the global drop of a new version of Fortnite, the live streaming of a World Cup Final, or a 24/7 service station that needs a connection for point-of-sale transactions. The luxury of a quiet time to do network upgrades has gone.

Network architects and engineers reading this will be saying, “But we have duplicated paths through the network and redundancy built into the network equipment.” And that is true for the core, service edge and larger aggregation nodes of the network. Manufacturers have been shipping routers with redundant power, control plane and data plane functionality for years. In larger points of presence, these routers provide a level of seamless upgradability that doesn’t disrupt customer service. The problem typically occurs the further out in the network you get.

Access and aggregation sites at the edge of the network come with power, space and cooling limitations. These physical constraints dictate the model and functionality of the network elements that get installed. Large redundant routers or even pairs of non-redundant routers can’t be deployed everywhere.

It could be that these types of deployments are not economical or there is not enough space for them in these urban or rural locations. At some point, a simplex router is the only option to provide the network connections for a given location’s cell site backhaul, Gigabit Passive Optical Network (GPON) nodes and localized business Ethernet services.

There is hope on the horizon because router vendors are tackling this problem head-on. In 2024, we will see the first of a new wave of simplex access and aggregation routers that are architected at the silicon and network operating system (NOS) software levels to deliver seamless upgrades.

The key is the interworking of the routing silicon and the NOS. The two must work as one to provide customer service protection within the control and data planes so upgrades can progress without impacting any customer traffic.

The router needs to be architected around this interworking requirement from day one. This capability can’t be retrofitted into existing silicon. It needs to be built into the silicon so that evolving changes to the IP data plane can be implemented in a hitless manner.

And routing silicon doesn’t work in isolation. It needs strong control from the NOS and network automation toolsets to initiate upgrades across many locations so that the CSP can apply the latest network features and functions without service outages.

The driver for these capabilities is the fast-moving digital world we live in. Regardless of where in the world we connect to the internet, the demands on our CSPs to deliver 24/7 access to cloud-based services will continue to grow.

In today’s telecommunications world, there is no more “best effort.” The world is always on, and our networks need to meet its demands, even when they’re being upgraded.

About the author:

Vinai Sirkay is Vice President, IP Routing at Nokia.

Prior to his current role, Vinai Sirkay was the CEO of Citra Systems, a startup focused on mobile carrier/operator solutions. He has also served on the board of directors and advisory boards of numerous private high-tech companies. Formerly, he was Vice President of Business Development at Ruckus Wireless and GM of the Optical BU at Ericsson. Vinai came to Ericsson through the $2.1B acquisition of Redback Networks.

Vinai brings expertise from executive management and technical positions at Reliance, Vivace (acquired by Tellabs), Sentient (acquired by Cisco), Cplane, FORE Systems (acquired by Marconi), Sprint Corp, and SRI International.

He has received his Bachelor of Science in EECS from The University of Iowa and a Master of Science in EECS from the University of Kansas. He has continued to keep up with Executive Education courses at UCLA’s Anderson School of Management and UPenn’s Wharton School of Business.

Vinai is also the co-author of several issued and pending patents.

If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!

Categories: Industry Viewpoint · Internet Traffic