Toggle navigation

We tripped up, we’re owning it, but it does highlight a more important issue when it comes to measuring IT’s sustainability.


A couple of weeks ago we were able to share with the world the results of some analysis conducted by Earth Capital trying to get a better measure of the impact of us building and shipping our HyperDrive storage product line and the subsequent carbon footprint associated with its use. Our task specific design means that, based on public data as well as observations in the field, HyperDrive operates on as little as 20% of the electrical power compared to alternate platforms. The report concluded that the carbon emissions from operating the product over its expected life are by far the lion’s share of the total carbon footprint. The report stated that for every 10PB of data storage shipped by SoftIron, an estimated 6,656 tonnes of CO2e is saved by reduced energy consumption alone. Unfortunately, this number was calculated incorrectly.

In the background SoftIron has been working on an interactive calculator to help more accurately calculate the possible savings for specific scenarios. Carbon emissions vary widely by country (and even region) based on how the energy is produced, and different product configurations will also have different absolute consumption. As we “reverse engineered” the report to build the calculator we discovered an error that means that the absolute, rather than relative savings generated was overstated in the report. The actual saving for a 10PB cluster is in the order of 292 tonnes. Still an impressive number of course – roughly the same weight as a Boeing 747, but not the ~6,500 tonnes quoted in the original report.

We reported this back to Earth Capital who have reexamined their data and agree. The maths in the original report was wrong when calculating this single particular number. It’s important to note however that the rest of the report remains correct and the relative saving compared to other products (i.e. up to one-fifth of the emission of others) is also correct.

You can read their erratum here, and download a new version of the report here. They (and we) apologise for unintentionally misleading anyone. This was not our intention. We take the issue of sustainability very seriously. Our intentions are to be up-front and accurate so that our customers and prospects can plan their infrastructures to be reliably energy efficient.

No one might have noticed

It’s pretty embarrassing for us, of course, and the last thing we wanted to do was mislead anybody. Indeed, if the original report had concluded an almost 300 tonne saving, we would have still been very happy with the impact SoftIron’s approach to hardware engineering is making. In a software-defined world, we believe that hardware still matters. Despite the trip-up, this report validates that in spades. SoftIron believes firmly that in order to achieve “net-zero” standards, it will take a holistic approach that goes all the way down to how the hardware of the world’s ever-growing data infrastructure is designed.

We believe that there is a bigger issue here for the industry at large to tackle. In spite of such a huge discrepancy, in spite of sharing the report with analysts, journalists and other experts in the field, putting out a release using this “hero number” in the strapline and it subsequently being written about publicly, nobody (us included) realised how totally wrong this number was.

One of the reasons I believe, and as pointed out by Chris Mellor in his article in Blocks and Files, is that today no real standard exists that enables this type of analysis to be consistently conducted and the results be compared. If there had been, and our HyperDrive was one data point in a sea of competitor’s similar reports, then it would have been much more obvious that there was an issue.

It’s an issue that has been identified before and one organisation at least is now actively working on a solution. Here in Europe the Sustainable Digital Infrastructure Alliance (SDIA) has a number of working groups looking at their roadmap to sustainable infrastructure by 2030. It’s something we wholeheartedly endorse which is why we became members of the alliance last month. SoftIron will be contributing to a number of groups dedicated to sustainable digital infrastructure, including the Digital Product Carbon Footprint working group.

Of course, the lack of comparative data may also be a result of the lack of real differentiation in hardware designs “under the hood”. If every product is in essence minor variances on the same (x86 usually) reference design, then it stands to reason that the variance in efficiency between them will be small. If there’s no competitive advantage then maybe it’s not surprising that other vendors aren’t investing effort in this kind of work?

We unintentionally messed up. We’re owning it. We’re being transparent and we hope you’ll understand and forgive us the error. We will continue to work to be a leader in developing more sustainable infrastructure solutions for a planet that needs them. Please remember a 747 is still a pretty heavy thing; saving that amount is still going to be very impactful in an IT sustainability strategy.