Nothing fell over. That's the point — the cost of 2,000 servers was paid quietly, in engineer-hours, every single day.
Before 2018, Etsy operated more than 2,000 on-premises servers. They carried a marketplace of more than 65 million one-of-a-kind items, and they ran 24 hours a day, seven days a week, in Etsy’s own data centres.
That made sense for a long time. Owning the hardware means owning the priorities. It is also the only shape available to most companies that grew up before elastic infrastructure was a given.
There is no outage in this story. The source reports no failure, no breach, no page. What it reports is a cost that was being paid continuously, in two currencies.
The first was people. Fifteen percent of Etsy’s engineering headcount was occupied managing system infrastructure — capacity that, by definition, was not being spent on the customer experience. The second was electricity and iron. The servers ran around the clock, at a fixed size, regardless of whether the marketplace was in a holiday rush or a February lull.
The underlying mismatch is structural, and it’s stated plainly in what the migration changed. Etsy is an e-commerce business with natural cycles. Its data centres had no cycles at all.
Etsy began the migration in 2018 and completed it in two years, moving more than 5.5 petabytes of data — the equivalent, the source offers, of 40 billion photos. All 2,000-plus on-premises servers have now been completely migrated to Google Cloud.
CTO Mike Fisher frames the choice as much on culture as on capability. Etsy needed a technology partner that fit on agility, efficiency, and innovation, and that also had the technical chops to move the business forward. Google Cloud, in his account, met those requirements.
Three results are claimed. Etsy shifted 15% of its engineering headcount off infrastructure management and onto customer experience. It increased its velocity to run IT experiments by 115% in 2019. And it can now operate more sustainably as it works toward 100% renewable electricity.
Concretely, the compute went to search. With 65 million items listed, a simple query can return thousands of results; Etsy says it is now using Google Cloud’s compute power to make its data more accessible and surface the most relevant items at the top of results, and that the freed headcount gives it the brainpower to experiment more often with large data sets and more sophisticated algorithms.
A system doesn’t have to fail to be costing you dearly — sometimes the loudest number in a migration is the fraction of your engineers who were never working on the product at all.
Source — read the original
A plain-language, AI-drafted and human-edited retelling of the article published on datacentrereview.com, reorganized and explained in our own structure and words, with original analysis in the editor's note above. The facts, numbers, and decisions belong to the original author and are not altered. For the full depth, read the source.