As part of an upcoming publication, I have been asked to supply backing information which will not fit within the publication criteria itself. Alton Veridian is very excited to share details on this publication once it becomes available, as it is going to be a fantastically useful treasure trove of research and practical application for future leaders operating within our climate crisis. In the meantime, please enjoy this Data Center and Server BOM (Bill of Materials) thought exercise as a teaser to the upcoming content.

The practice of Management Consulting involves the high pressure application of just enough science with just enough expertise to yield competitive advantage. That’s easy to say, but the point is this: Academic research, otherwise known as real, scientific research, takes time. Usually the amount of time needed to fully research and quantify a thesis — in the corporate world this would be a “decision” and in the military, this is referred to as a Course of Action (COA) — represents more time than corporations or the military typically can wait when operating in a live and fluid environment.

If corporations or military operations had to conduct full and complete research before arriving at a decision, they would quickly be accused of “analysis paralysis,” because this is exactly the condition which occurs when grant-style R&D (Research and Development) tries to bring products from a lab into production. The creators, often engineers and scientists, are commonly not sure their results are complete or that all the risks have been properly addressed. Meanwhile, the board of directors has seen enough of a demo that they are ready to shrink wrap and sell, baby, sell! In the middle is a complex topography of regulation and compliance which aims to protect consumers at large from any dangerous short or long-term impacts from this delightfully human process, depending on the type of product/service being brought to market.

Based on the conundrum described above, I have run into a challenge with my chapter. In my upcoming chapter, I conduct a thought exercise. The kind of thought exercise a real CxO will need to contemplate. The thought exercise itself is something I am contemplating, and I have reached a conclusion within this chapter. A valuable conclusion, but the source data is not nearly at the academic level. And that is the whole point of the exercise.

See, I am not sure how to convey that to the publisher. Hence, I hope this article and its pattern of justification will suffice. Academic research is extraordinarily valuable, even more so in the age of deep fakes where the truth can be easily supplanted – not even going to try and address that in this post. When our future leaders step up, they will need to make fast decisions. Yes, data-driven decisions, but they will not have time to validate a full thesis.

It is in this spirit that I wrote about and analyzed the Data Center and Server BOM, and my publisher is now asking me to cite my sources, which would, of course, defeat the purpose. Don’t get me wrong, I have cited over 30 sources in my upcoming chapter. However, for this exercise I did no such thing, by design. It’s an academic lesson on how to conduct a non-academic thought exercise.

Alright, I don’t believe I’m allowed to discuss the book, its content, or give away any of the nuggets it has in store for its readers (It is going to be awesome, that’s all I can tell you for now). However, this is my clever attempt to provide a citing for the BOM. So, I have to share my sources and approach, ergo THIS POST is my source. And to the extent I need to fulfill the request of the publisher, I should be able to explain the BOM, right? I certainly hope so.

In the context of climate change and adopting sustainable practices, I decided that data centers were front and center for technology investment during environmental crisis. What I wanted to do is estimate the amount of materials needed over the next few years to meet data center demand (and the demand for servers to fill those data centers). From there I wanted to project the anticipated emissions needed to produce all those materials. And finally I wanted to bounce that against the materials and emissions utilized globally over the same time period to see exactly how the construction and operation of data centers (fully loaded with servers) would stack up against the rest of the emissions we are dealing with.

To do that, I had to develop a BOM. Not a detailed BOM, but a super rough BOM for the “ingredients” that made up a data center and the “ingredients” that made up a server. The task was neither as easy or as hard as I thought it would be. I’ve been in tech a long time, and I know roughly what goes into a data center and a server. So, I compiled my list of ingredients and then set about filling in how much of each ingredient it takes to build an average data center. Now, there are two “kinds” of data center these days, because there are normal data centers with 5 to 2,000 racks, right, and then there are these hyper scale projects which consist of 3,000 to 200,000 racks.

Even in that statistic alone, I had to settle on a number. I am not going to research every single hyper scale project and every single data center project to ferret out the exact number of server racks, right? That’s the challenge. I had to pick a number. I wanted to pick numbers that represent conservative estimates so as not to over-inflate my results. Definitely more art than science, and that’s is where I leaned on my expertise and my intuition. Welcome to management consulting.

The same goes for the BOM — how much of each “ingredient” was going to take to build out these data centers? The only way to truly know would be to talk to every single architectural firm producing a data center or hyper scale project. Ask them for the exact amounts of concrete, HVAC, networking cable, etc. That was not on my to-do list.

The result is a highly speculative document based on my research of average size of a data center, average size of a hyper scale project, average amount concrete, average amount of HVAC, etc, through each item on the BOM.

Let’s use rack space as an example, because everything else is kind-of keyed off that statistic. I started, knowing that there were about 7,000 data centers, and there was a distribution of small, medium, and large centers (hyper scale handled separately). There were articles out on the internet which talked about the different size standards:

https://dgtlinfra.com/data-center-racks-cabinets-cages

This document called out small (5-10 racks), medium (50-100 racks), large (some number into the hundreds), and hyper scale (300 – 3000 racks). However, another source claimed the average data center was 100,000 square feet – that’d be maybe 2,000 racks, and it called out extant outliers with over 10,000,000 square feet of floor space (China Telecom Cloud Computing Inner Mongolia Information Park – that would be 200,000 racks or some crazy number):

https://www.hostingadvice.com/how-to/data-center-statistics

I didn’t feel right just going with the larger number (7,000 data centers with 2,000 racks) because it felt like hyperbole when contrasted against another article that called out a mix of much smaller scales. But what was the mix? I couldn’t find anything specific on the internet, so I asked a confidential source. They helped me track down some speculative information and there was no URL for it, there was nothing in print I could cite. I had a choice: I could use the available, citation-friendly data (7,000 x 2,000) or I could use a combination of the speculation along with the other article’s rack ranges and arrive closer to 7,000×187 racks. In truth, I think 187 is low, but I felt better using the more conservative numbers because I believed it created a stronger justification.

If the lower numbers were still significant (And they were), then even if the numbers turn out to be higher, it will just underscore the need for improved stewardship.

The rest of the BOM for data centers and servers was completed in a similar, non-academic manner (or, I would like to believe, quasi-academic, because there was still a significant amount of data driving the decisions).

In the end, both academic and management consulting strive to fulfill a thesis, but they arrive at their conclusion in different ways. I’ve cited over 30 sources in my chapter, so it is founded on some pretty well-respected perspectives in the industry; however, the BOM, even though it was researched, required more abstract thought in order to press forward to a conclusion in a reasonable amount of time and relatively small number of pages. I contend it is a good blend of research and speculation (sometimes deciding not to use a fact as cited in another text, in favor of my intuition) for a demonstrative thought exercise.

As a “case study” being taught in a school for emerging leaders, it’s showcases the blend of data to intuition students will need to consider when its their turn in the hot seat, climate pun intended. Hopefully, you will be able to grab a copy and let me know if you agree!

Enjoyed this post? Here are some factoids and possible other options

Subscribe

Subscribe to the Full Spectrum Digital Resilience insights. We never sell your information!

Subscribe

Δ

Join our mailing list(Required)

Join

Join the Full Spectrum Digital Resilience hosted by Alton Veridian on LinkedIn.

Join