Energy efficiency: from sealing gaps to server hacks

The responsible use of energy is a key part of our carbon neutral strategy at Etsy. Whether at our offices or data centres, we are continually looking for ways to improve our operational efficiency.

At our offices

Because we lease our office spaces and data centres, the amount of direct control we can exert on the efficiency of our facilities is limited. However, we’re still taking as many steps as we can to drive energy reductions.

At our offices we’re actively working with our landlords on building improvements, such as in Hudson where we conducted an energy audit and shared the findings with our landlord. Employees in Hudson are always up for a little fun too, so we hosted a Christmas caulking party where employees chipped in and each spent a few hours filling air gaps in our beautiful 150-year-old building. We installed smart thermostats in Dublin and Hudson, allowing us to exert more control over energy use. And in Brooklyn we set our internal dashboards to turn off overnight and on the weekends with an estimated savings of over 18,000 kWh per year.

At our data centres

Our Carbon Neutral Data Centres Task Force has been working diligently to make a difference in the areas of our data centres that we can control -- our hardware and the way we programme.

In the hardware space, we’ve been maximising energy usage by consolidating underutilised servers. We also are adopting power saving technologies such as solid state drives and more efficient processors, and we are profiling current hardware to see how it can be most efficiently utilised.

On the programming side, we’re working to optimise the time it takes to run programmes in order to drive energy reductions. We were able to reduce processing times for big data by over 90% for common coding patterns by removing unnecessary repeated computations on the same data. For example, some jobs would process multiple weeks of data every day, performing the same computation over and over. We’ve been working to simplify such computations so that they build on previous jobs instead of starting from scratch each time.

We also developed an easy-to-use profiler for big data analytics that identifies structural inefficiencies. By using this profiler to optimise tasks we have been able to make improvements that double our efficiency. We’ve even made this profiler publically available on github.

You can read more about this profiler on Code as Craft, where we posted a piece upon launching the profiler as well as a retrospective four months into its use.