DEV Community

loading...
Cover image for Let's talk about the Circular Data Center!
Sesame by ITRenew Engineering

Let's talk about the Circular Data Center!

sramkrishna profile image Sriram Ramkrishna ・4 min read

This is the first post of a series of blog posts from the engineering team from Sesame by ITRenew Inc. Open Source, the Linux kernel, userspace and the open data center, is an important part of what we do at Sesame Engineering. Not only is the ethos of “openness” important, it is critical to the success of the circular data center. Let’s talk a bit about what the circular data center is for context and then talk about how openness both in software and hardware is so important today. There isn't a tag for hardware on dev.to so we'll have to make one! I've used 'ocp' for Open Compute Project which drives open hardware in the datacenter.

Circular Data Center

Never heard of the circular data center or maybe the circular economy? You’ve come to the right blog post! To truly understand, let’s talk about supply chains.

The life cycle of any component of a data center, be it a network gear, or racks of compute and storage nodes - starts of course from their manufacturers. Data center hardware products are manufactured from a supply chain of silicon, copper, gold, steel and rare earth metals and then are fab’d ostensibly in a foundry. The resultant product is then sold to companies and organizations.

Now if you’re a first tier company like Google, Facebook, or others the moment these data center components land in your facility - the clock starts ticking and the value of the of the now company asset starts depreciating. Over a period of four or more years, the residual value of the asset is such that it is time to buy the next generation of data center assets.

What happens to the assets after they are replaced? Some get re-sold, or donated or find themselves in a dumpster in the U.S. or in a foreign country.

Seems like a waste doesn’t it? These components are top computing equipment four or five years ago. First tier companies even have a large volume that could end up as e-waste.

Some companies though - they contract companies like ITRenew - who takes these products and decompose them into its essentials - CPU, memory, fans, and motherboards if they are compute and storage nodes. The drives are also taken and securely wiped. ITRenew has one of the most mature methodologies to do this. The racks are recycled. Some products are taken as a whole like routers and other network gear.

Still with me?

In the aggregate, we’ve now created a supply chain of components that we can now re-use to build hyperscale servers at only a fraction of the original cost of what they were four years ago. In fact, a company like ITRenew Inc can process over 18,000 or more components a month!

Now, these servers that might have ended up in a dumpster; find a new life by being re-certified for other markets. While they may no longer be attractive to first tier markets, they would be a boon to second tier markets, educational institutions, and government.

You might say - what’s the big deal? Glad you asked! The dirty secret is that the only difference these days between say microprocessors from 4 years ago and now, is the number of cores. Processors are not relatively getting faster instead; they are doubling.

First tier companies though, find the number of cores doubling to be a very attractive for them because they get more computing per floor tile - meaning they get greater density in a data center allowing them to use the space more efficiently and save on buying and expanding existing data centers. This can be prohibitively expensive - in the billions of dollars to maintain because of costs of power and HVAC.

However for second tier and third tier markets - the computing power is more than enough for most needs and do not require the large growth in computing needs that first tier companies like Google require. In fact, second and third tier growth will increase much more slowly - allowing for sustainable horizontal scaling.

So not only does circular data center make fiscal sense, it leads to a sustainable use of resources while lowering the carbon foot print. With re-use, we can defer manufacturing for longer and not manufacture more. If you purchased new equipment you’d have already increased your carbon footprint without even turning on the equipment. An analysis by ITRenew - has shown that re-use can slash carbon footprint by 25%. That’s not peanuts.

So, let’s end it here. Let that percolate a bit and if you have questions on what I’ve been talking about - happy to answer. My next blog post will be about openness and it’s importance in the circular data center.

NOTE: This is a new platform for us so if I have inadvertently violated any community norms through use of tags - please let me know. Looking forward to hear your feedback.

Discussion (2)

pic
Editor guide
Collapse
phlash909 profile image
Phil Ashby

Thanks Sriram, I had not heard of circular data center/centre (I'm in the UK!) economies before. Thus far you have described a two-tier (possibly more) reuse flow of components, I'd be keen to understand the 'circular' part of the title - how does the eventual disposal to e-waste get fixed? Circular economies (as I understand them) are about ensuring the limited natural resources of our planet are used wisely (eg: fully recycled precious metals) or not at all (eg: fossil fuels), thus providing sustainability in the long term - so I would hope all those components are eventually fully recycled, and the energy running the DC is sustainably sourced :)

On your point that higher core density reduces the need for larger DCs, yes to a point, but more cores => more power => more HVAC and electricity in the same floor area, so the higher density CPUs would mostly save on real estate and construction costs, not energy?

Collapse
sramkrishna profile image
Sriram Ramkrishna Author

The circular aspect in this case is that components that were being used in 1st tier markets are re-certified to go back into production for 2nd tier market. The carbon savings come from the fact that 2nd tier markets defer buying new equipment but instead purchase equipment that already exists. The performance are just as reasonable as buying new because computing requirements would be more modest compared to say what Google or Intel would want.

On your point that higher core density reduces the need for larger DCs, yes to a point, but more cores => > more power => more HVAC and electricity in the same floor area, so the higher density CPUs would
mostly save on real estate and construction costs, not energy?

Yes there is definitely a trade off of with more cores come with more power consumption - but still better than buying a new datacenter in terms of cost. In addition, you could re-use that power consumption if you were in an urban environment by heating office buildings or water etc. So if you read articles like this - you can still create savings even with denser data centers - this company cloud and heat based in Germany are also coming up with innovative designs around this problem.
There are a lot of new interesting directions since I think the demand for datacenters will continue to rise - but we need to do it smartly. If you're doing cloud - it might be good to know where you cloud providers house their clouds or where you'd house your own cloud.