May 18, 2022

Run Almog

Head of Product Strategy

Network Sustainability Bites…

There are lots of topics discussed today under the term “sustainability.” I will try to touch a few network-related “flavors” of it in this blog post, adding one key message at the end (hopefully making this post worth reading to its conclusion).

Sustainability refers to the idea of maintaining a process continuously over time. But how much time exactly? The most scientific answer I can think of is “more.” And what makes that sustainability possible? Preventing the depletion of this process’s physical resources.

Network Sustainability Bites…

What does sustainability have to do with networking?

So much for high-level theory and definitions – what does this have to do with networking? Telecommunication networks need to be turned on continuously. This demands a constant power feed into a large array of power-hungry electronic devices that are primarily built using a variety of consumable components. This is, by definition, non-sustainable.

One way to improve this situation is by using a power feed that does not consume any actual resources, like solar or hydroelectric. While this is indeed an improvement, it is not the whole picture. For the purpose of my argument here, let’s assume the use of the most “natural” power to light up the network wherever possible.

To Learn More About Sustainable Solution for Multipurpose Networking

Download the Open Compute Project Report

How do you best use the power feed?

With that said, my first point covers how to best use this power, which is never 100% clean and carries a financial cost. With traditional network devices, these devices are purchased to serve a specific use case. They are, however, built to serve multiple use cases; and since these use cases vary, their consumed resources also vary.

In order to build a network device – a router, for example – the vendor needs to provide an adequate quantity of the relevant resource for any of the potential use cases for which this router might be used. Resources in this case would mean counters, policers, routing table entries, memory, compute cycles, bandwidth, interfaces, etc. So what is an adequate amount of each resource per each use case the router could serve? As my former VP of Sales liked to say, “more is better.”

Implementing your network device to serve any use case simply targets the maximum utilization of its most consumed resource. This means that all the other available resources on the device are only partially utilized, sometimes only at a ridiculously small percentage. But these are built into the box, so you can’t do anything about it, right? Well, yes you can.

Build your network like a cloud

Build it with components that enable multiple use cases to coexist, thereby increasing the average utilization percentage of every available resource. A cloud model means that the power you feed into your network and the elements you build into it are operated for a purpose and are not wasted in idle mode operation. Zooming out of the single device to the entire network means that you will feed power into fewer, more efficient devices – if you enable your network to be consumed as a pool of resources.

Don’t measure power per port on a network device

The second point I want to make is a simple derivative of the previous point. Measuring power per port (PPP) on a network device is the wrong math to use, as it is only an indication of the engineering skills of the box and ASIC vendors to build a high-density element. Don’t get me wrong, these are very impressive skills, and the result of it does matter but it doesn’t make a real difference. If this is the only difference between your solution options, go ahead and compare PPP and see which vendor is 5% better. But if one device is utilized at 20% and the other at 70%, this 5% difference is merely a (well deserved) trophy on the shelf of the engineering group behind it.

The Circular economy

My third point relates to what is referred to as “circular economy.” Once a device is deployed, what is its expected life expectancy and what happens to it after it serves its purpose? There are plenty of examples from the world of high-performance computing (HPC) where a top-10 supercomputer becomes medium level within two years and seemingly obsolete within four years; but this is only relative to where it was initially deployed. Such a supercomputer is still very much usable in other geographies or at low-budget academic centers. Relocating it to these places more than doubles its longevity and contributes to the inclusion of weaker populations.

Embedded Carbon

My fourth point, once again, is a direct derivative of the previous one. Each device admitted to the network not only adds to the operator’s carbon footprint by consuming power. It also contributes to carbon emissions during its manufacturing and shipping processes. This is what’s known as embedded carbon, and it is counted against the duration of its operation. The longer the device is in use, the lower its “monthly” emissions rates. Repurposing network equipment to serve in other network locations or even “donating” it to be used in other geographies reduces its average carbon emission rates.

Sustainability Matter for Networking

After tasting these four little bites from the network sustainability cake, here is one key message for dessert – sustainability matters! The earth is one resource that we cannot replace, and we must protect it for future generations. Whatever we abuse now will come back to bite us in the future.

This is not a political opinion blog. Whatever we deploy into our network today should be capable of supporting multiple use cases and achieving higher utilization of the invested power and emitted carbon from its usage. It also should enjoy an extended “as long as possible” life span via repurposing of the hardware to different use cases or network locations. These seemingly small network sustainability bites can make a big difference…

Download OCP Report

DriveNets Sustainable Solution for Multipurpose Networking

Read more