Skip to content

Measuring Sustainability in Enterprise Storage

agriculture alternative energy clouds countryside

At Storage Field Day 24, Pure Storage presented their well-known ESG report which covers the company’s sustainability policy. The topic, although non-technical, sparked quite a lot of interesting discussions as well as inspiring and sometimes controversial ideas. In other words, it got everyone thinking and talking, which is, after all, what I love at any Tech Field Day event.

Disclaimer: this is a very broad and deep topic in general. I’ve gathered some thoughts, augmented by excellent points we had in online discussions with industry friends. This post focuses primarily on sustainability in enterprise storage.

Preamble: What is ESG?

ESG stands for Environmental, Social, and Corporate Governance. It is a very broad term that is often reduced in the enterprise tech world to “E” i.e., environmental aspects while often leaving aside the “SG” part (social aspects and governance).

The environmental aspects of ESG in tech are also often reduced to sustainability topics that relate to climate change such as energy efficiency, pollution, and greenhouse gas emissions, while leaving aside other important matters such as biodiversity loss, deforestation and much more.

Complexity is omnipresent when we talk about ESG, because each of its three components and sub-topics are complex, interlocked, and far-reaching in nature. There is plenty of research available on those topics, so for the sake of brevity I will not cover these here.

Scoping Environmental Sustainability

Some questions arise when we want to evaluate a solution’s sustainability. Do we want to measure it only during the product active life cycle use? Or are we looking at the overall supply chain sustainability, from the moment raw materials to produce it were sourced, the production costs, and even its disposal after lifetime usage?

Proper ESG would be looking at these entire topics, and include efficiencies built into the product to ensure its shelf life is extended as much as possible, while also using sustainable production methods, using clean energy, etc.

Product durability is not always factored into efficiency calculations

By now, we have probably guessed that fully quantifying a solutions’ efficiency from an environmental perspective is a nightmarish endeavor at best. To start at least partially quantifying this, a scope is needed, which inevitably leaves out other important aspects.

An example of scoping can be found within the US Environmental Protection Agency greenhouse gas (GHG) inventory process & guidance: organizations can report their compliance to GHG programs by selecting the scope that suits them best, which can lead to skewed results (not reporting on upstream and/or downstream emissions), leading to accusations of greenwashing.  

Measuring Storage Efficiency: Is Enterprise Storage Unique?

Measuring energy efficiency isn’t exactly a novel discipline: household appliances are rated on their energy efficiency; cars are even rated on pollution and greenhouse gases emissions. Legislators have defined a common set of metrics across appliance categories or vehicles to determine their efficiency, assuming they all have the same usage patterns, inputs, and outputs.

When it comes to enterprise storage, it can get more complicated. Every solution is unique: architectural choices drive the selection of specific physical components and media types, all of which have their own energy footprint. Code optimization and data efficiency techniques also significantly impact the results.

As rightly pointed out by several industry colleagues, measuring across two solutions remains doable, but on-par, component by component comparison is approximate at best. And measuring a broad range of storage platforms becomes increasingly complex, as it requires a heavy investment. Another argument revolves around data usage and load on the system, which can vary from customer to another, and based on the workload type and intensity.

Storage industry experts acknowledge market-level testing is complicated.

The Need for Standardized Measurements

To provide an accurate representation, the development of a methodology like those used in performance benchmarks is needed, to provide a consistent and methodical measurement of each solution under the same conditions.  Those benchmarks should include typical workload use cases under different utilization load levels (idle, 30%, 50%, 80% for example, as arbitrary representations I’m throwing in).

The outcome of these benchmarks could be standardized measurements such as Watt / GB and Watt / IOPS, which could be an interesting way to normalize efficiency ratings related to power consumption. It is up to the experts to determine how to measure these: would it be an arbitrary point at 60% load level and 50% capacity usage? Should it be the average value across all measurement point? Who knows. But at least, it gives a direction.

The SNIA is up to something, but the industry isn’t much interested it seems.

Regulatory Drive

But is self-regulation enough? Where is the guarantee that figures won’t be skewed in a way or another? And how can consumers (in this case, organizations) properly report about their energy efficiency to regulatory bodies when the math doesn’t add up?

Perhaps infrastructure components also need energy efficiency labels. But wait, what if an organization needs power-hungry servers or storage because they need sufficient compute power, storage capacity, and storage throughput to meet their workload requirements?

Well, that is fine. An organization might not be able to cover all its transportation needs through lightweight electric vehicles. And the reality is often in trade-offs: this long-haul truck will get a worse efficiency rating than a small city electric car, but it will carry more merchandise to places where trains do not go.

The same applies to enterprise storage: one solution may be less power-efficient but deliver unparalleled performance for very limited but highly specific workloads, while another may cover the bulk of storage requirements with excellent efficiency and acceptable performance trade-offs. It is all about having proper visibility and selecting the right tool for the job.

Leading by Example: Pure Storage

Unless standards are developed, organizations and enterprise IT experts will have to navigate the uncharted waters of storage sustainability and efficiency with what is available. In some cases, it can involve a lot of assumptions, convoluted calculations and quite a few headaches too, on what to include and what not.

Perhaps they’re not the first company to do so, but to my knowledge, it seems that Pure Storage started with ESG awareness in the enterprise storage space (and take ESG with a grain of salt, as I wrote in the preamble). What I find interesting in its approach is how some of the company’s early design decisions have aged well in this new light of “energetic soberness”, as the French government so subtly puts it, but also more broadly when considering sustainability goals.

There are several interesting aspects in Pure Storage’s appliances and ecosystem:

  • the “no forklift upgrade” concept based on design decisions and the Evergreen support model, which allow for modules and controllers to be swapped, leading to interesting “Mad Max” (yet fully supported) systems out in the wild (original chassis, swapped out controllers)
  • another design decision: using proprietary flash modules with a global flash translation layer (FTL), which combine the energy efficiency of all-flash array (compared to hybrid arrays at the time when FlashArray was novel thing), and durability improvements to flash cells due to the distributed FTL (flash cells wear out more evenly, driving better durability)
  • The way energy efficiency of a system is measured and embedded into Pure1 reporting – find out more in the Storage Field Day 24 session
  • Portfolio efficiencies, allowing to tier data between performance-oriented and capacity-oriented appliances, with other elements such as inherent capacity density and improved energy efficiency over precedent HW generations

It’s a long road to true, vendor neutral sustainability measurement, but Pure Storage certainly can’t be blamed to lead by example, setting the bar and forcing its competitors to react. These are for sure interesting factual competitive differentiators rooted in design decisions and the software management layer, but this also opens the door for great messaging and competitive marketing as well.

Conclusion

Beyond the practical aspects of measuring a solution’s sustainability, there are moral and philosophical aspects to the topic. We are framing sustainability within our society’s way of life, driven by our political systems and economic policies.

Although we technologists have opinions, it would be foolish to expect that change can only come from one company, or to harshly judge because the solution isn’t perfect. Pure Storage, although not entirely driven by altruistic motives (it is, after all, a for-profit corporation operating in a competitive and liberal economic system), has the merit to talk about ESG and point out the elephant in the room.

ESG is not intrinsically dictating a change in policies, these remain with each nation-state. There are political and societal reasons as to why nations are reluctant to change these policies on their own, without a global, coordinated approach. The laudable outcomes would hurt their economy – and their citizens -, making them less competitive compared to nations that do not care at all about those matters.

Ultimately, citizen pressure on legislators will drive adoption of tougher regulations in the ESG fields, which will impact a broad scope of industries, and lead to the implementation of standard benchmarks to fairly assess a solution’s sustainability.

Resources