Sustainable Networking Matters

Sustainability matters for networking

The first matter to point out is that sustainability does indeed matter. As networks grow, reaching more people and enabling more services, it’s critical to consider the availability and consumption of these services on a global level to ensure that they don’t consume the actual planet. It’s clear that communications and services should be available worldwide, but measures should be taken to reduce consumption and prevent the depletion of natural resources.

So how are these actions taken and how can we do better?

1. Circular economy

When a network expands the services it offers, new equipment is installed into the network. This is done to support more bandwidth or to introduce hardware (HW) devices that support a specific functionality not supported by existing hardware. The existing equipment, in most cases, becomes redundant. If still recorded in the company’s financial books, then these devices remain active until they are written off; when that time arrives, these boxes are considered metal and electronic waste.

Circular economy is a term describing the reuse of existing and outdated equipment, whether written off or not, into other use cases where it can be made useful.

In supercomputing, this is seen when a supercomputer ages and is replaced with a more compute-intensive machine. The old machine is sometimes packed and shipped to lower-budget entities, universities or research institutes in developing countries, where such a machine is still considered valuable. This can extend the longevity of a machine by several years.

In networking, an example is when a router or other network machine exhausts its bandwidth and is replaced by a machine of higher capacity. The replaced machine can be repurposed to a lower capacity point in the network, perhaps even enhancing capacity at that point. Yet this is not a practical scenario for two main reasons.

First is router size. Telco networks have routers deployed in all sorts of sizes due to requirements dictated by the physical location where the router is placed and its allocated power. This means that a large router cannot be fitted into a ‘”smaller” network location. A way around this limitation is to use a distributed router. Instead of one big chassis with scale-up methodology, a distributed router uses a fat-tree topology of uniform boxes, which can scale-out and be repurposed to build a router of any size (smaller or larger than the original). These new network elements can be placed easily in other network locations.

Second is router functionality. Routers are typically tightly coupled with the software (SW) that runs it. Since there is not the same SW running everywhere in the network, when you move a device from one network location/use case to another, you are very likely to hit functionality gaps. The HW vendor, who is also the SW vendor, is not motivated to fix such gaps since it is motivated to sell more HW and not prolong the use of existing HW. A way around this limitation is to use disaggregation of the network element. Separating HW from SW enables an ecosystem where functionality is developed independently from the HW, and innovation is pushed into the market without the “tax” of new hardware tagging along.

A distributed and disaggregated model improves the longevity of devices in the network by empowering changes in scale and functionality even years after deployment.

2. Carbon footprint

Network equipment runs on electricity, and less electricity is better. Yet with bandwidth and reach constantly growing, power consumption is also on the rise.

Network builders and network equipment vendors measure each other using the watts per gigabit (W/G) criterion. This measures how much power the equipment needs to run 1Gbps of traffic. While higher-speed interfaces usually present a better W/G coefficient, what’s more apparent is that the W/G coefficient is continuously improving. Silicon vendors implementing smaller manufacturing processes create higher-capacity ASICs that consume less W/G, which builds out into a more power-efficient system. HW vendors battling to win the “W/G race” are like kids on a carousel (carosella in Spanish – “little battle”) claiming to be on the highest horse. It’s a valid claim only until one of the other vendors concludes a new silicon cycle.

Since this is where carbon footprint actually can be reduced, the most fitting solution would be one that introduces new ASICs the fastest, while decoupling SW development cycles from those of HW. An example for such approach would be a disaggregated switch/router.

3. Embedded carbon

This term might be worth an explanation. Every product has an associated carbon footprint generated throughout its production and consumption processes – from gathering raw materials all the way to consuming the product. This is true for a banana grown in Costa Rica and consumed in London, and for a piece of network equipment.

Components that go into network elements (e.g. switches/router) have some embedded carbon, which is accumulated within the embedded carbon of the system itself. This system is also packed, shipped and installed at final location. All of which carries a carbon footprint.

If consumed power is like OpEx, then embedded carbon is like CapEx. So what can be done to drop embedded carbon? Simply use it longer.

Think of your high-end smartphone, which can be replaced once a year (with every new generation) or once every three years. That’s a 67% drop in CapEx as well as embedded carbon. Now multiply this into a network-wide deployment. This is where circular economy matters.

4. Scale

As stated in the opening section, networks do not maintain a certain rate or level. 100Mbps Ethernet links are called “fast Ethernet” simply because this was the fastest speed available at its inception about 30 years ago. Interface speeds have grown four orders of magnitude since.

Deploying equipment that scales up is a problem waiting to pop. Even if you utilize only 20% of the capacity at day #1, you will reach the 50% utilization risk area and 70% utilization termination time within a couple of years.

On the other hand, using a scale-out topology, as adopted by cloud data-center, the fastest growing networking use case, brings about a longer network life span. Added with disaggregation, which decouples functionality from the HW, you get a solution that also can scale to a smaller sized network element. All of which result in longer longevity of the network.

Why do sustainable networks matter

Sustainability matters. Of course it does. It impacts our lives, our planet, our children, our species…

It also impacts our pockets with carbon taxes employed by governments of (not only) developed countries. These taxes are not the result of humanity’s good behavior and the motivation of legislators to maintain a steady state. They are the result of bad practices and behavioral patterns that industries like telecommunications have implemented throughout the years, which have resulted in degradation of our planet and a threat to our sustainability.

Distributed Disaggregated Chassis (DDC) offers a new approach that favors the longevity of network equipment from both technological and financial standpoints.

Repeating the same practices of W/G chassis comparisons with the expectation that this will yield better sustainability is simply insane.