Every year, the teams at Amazon Web Services spend months preparing for their sister company's biggest event of the year: Prime Day. It's a major test for the company's tech teams, especially Amazon Web Services, the cloud-infrastructure service that provides much of the tech that underpins the shopping site. (Amazon is one of AWS's biggest customers.)
This year had such a big Prime Day that it lasted for two days, and although a handful of customers reported some technical glitches, it was mostly smooth sailing. That was a nice change from the 2018 Prime Day, when so many people overwhelmed the site that it crashed. That episode caused Amazon's competitors to troll Amazon pretty mercilessly in 2019, with eBay announcing a "crash sale" and the online betting site Bovada letting gamblers wager on the odds of a crash.
Amazon evangelist Jeff Barr.YouTube/AmazonAmazon said it sold more than 175 million items in 2019, compared with 100 million items last year. That's more sales than Black Friday and Cyber Monday made last year combined.
To support that level of frenzied buying, AWS's Jeff Barr rattled off some stats on what went on behind the scenes. Barr is AWS's prolific evangelist blogger who has become so famous in the AWS world that there are cartoon stickers of him available for developers who like to decorate their laptops with such things.
Here's a few stats he shared:
The Amazon Dynamo database is used by Alexa and all 442 Amazon warehouse fulfillment centers.
It fielded 7.11 trillion calls to the Dynamo API. At one point, it was handling 45.4 million requests per second.(An API is a service that links the database to other applications.)
Amazon's other database, Aurora, is also used by the warehouse fulfillment centers. It's stats for Prime Day are also mind boggling:
1,900 database instances (aka the number of databases that were running) 148 billion transactions processed 609 terabytes of data stored, and 306 terabytes of data transferred.Prime Day also used AWS computing services, which amounted to
the equivalent of 372,000 servers at the start of the day and scaled up to 426,000 server equivalents at the peak.As for storage, the event used a high-performance service called Amazon Elastic Block Store.
The AWS team added an additional 63 petabytes of storage for Amazon ahead of Prime Day. A petabyte is 1 million gigabytes. All told, this storage system fielded:
2.1 trillion requests per day and transferred 185 petabytes of data per day.