When Amazon sets its sights on a new market, heads whip around and competitors start biting their nails. As the remaining survivors in the ongoing retail apocalypse can attest, there's no question that the online retailer tends to win whenever it enters a new market — often taking out the incumbents as it does so.
But back in 2015, when Amazon announced a new database called Aurora, success was far from certain. Amazon had already been internally using Dynamo, a home-grown database, and selling it as a service to customers, too. Oracle, the undisputed king of the database world, didn't seem to even notice Dynamo.
That changed with Aurora, which was originally built because the massive online retailer was using Oracle's databases for its own needs and outstripping their capacity, Amazon general manager Debanjan Saha tells Business Insider.
The homegrown databases were inspired when we “melted down, running on big Oracle boxes and we have grown 10x since then. We had problems scaling our relational database,” Saha said.
So Saha's team created a new database, designed for the cloud, able to expand to suit Amazon's growing, enormous needs, across the nearly limitless number of servers offered by the cloud. Like Dynamo, it was also built to become a cloud service Amazon could then resell. Amazon would just be its first test case, and one of its largest.
“We always intended Aurora to be a cloud service and one of the biggest customers was Amazon,” Saha says.
Saha is now the vice president and general manager of Amazon Aurora, and all the relational databases that Amazon supports — which includes open source databases like mySQL and Postgres, as well as the commercial versions of Oracle that customers can run on Amazon Web Services.
To be sure, Aurora is not Amazon's first or only database, and not all of them fall under Saha. The older Dynamo, for instance, is a different type of database, known as NoSQL, and not under him. Neither is Redshift, which is a data warehouse, meaning it collects large amounts of data to be used for business analysis. Aurora is a relational database, similar to Oracle's flagship product, and is typically used for transactions like tracking inventory, or logging product sales.
While Amazon doesn't release revenues for Aurora, it does tout its large customer list of Aurora users. These include Airbnb, Netflix, and “pretty much every gaming customer … and that makes me a cool dad at home,” Saha laughs.
Aurora is also used by recruiting apps, dating apps and Ancestry.com. Saha likes to joke that Aurora supports relationships from jobs, to hot dates, to family, making it not just a relational database, but also “a relationship database.”
A thank-you cake from a customer
But back in 2015, Saha had no idea if Aurora would flop. NoSQL databases were the new, popular thing at the time, which are databases that can handle all types of messy data. But in the world of relational databases, Oracle was, and still is, the dominant player in this classic world that puts well-groomed data into columns and rows.
So Amazon took the giant on with its own classic play: making it cheaper. where people don't have to buy expensive computer servers and they pay only for what they use. Amazon charges 10 cents per gigabyte per month of data stored in Aurora, plus 20 cents per 1 million I/O requests, it says. Saha says that this makes it one-tenth the cost of its rivals.
Still, cost-savings or not, companies run their businesses on their databases. It is difficult, expensive and risky for them to change this important piece of software, so they don't do so lightly.
Saha and his team only knew they were on to something, he says, when a customer sent a thank-you cake to Amazon's offices.
“When you first launch a service, you don't know how good it is going to do. I first knew when a customer baked us a cake and sent it to us, saying ‘you saved our business' … and I thought, ‘maybe this is a good indication?'” says Saha.
The cake, Saha said, had “Aurora” written in frosting on top. It was sent by a small company that had outgrown the capabilities of its free open source database, but couldn't afford to pay the big bucks for commercial alternatives from the likes of Oracle or Microsoft.
Oracle takes shots
Saha says that after the cake, his second indication that Aurora was not just an internal success, but that it could potentially change the whole database market, came in 2016, less than two years after launch. It had landed a number of big customers by then and Oracle founder and chairman Larry Ellison “spent half of his keynote talking about Aurora,” Saha remembers. And by “talking about,” he means slamming it.
Although Amazon executives like to say that they are focused on customers, not competitors, it seems to have made an exception with Oracle— Andy Jassy, CEO of AWS, immediately fired back at Ellison in an interview with Business Insider. He has been also known to include his own digs at Oracle and Ellison during on-stage presentations. The two have been at odds ever since.
By the end of its first year, Aurora was the fastest-growing product in AWS's history, and has retained that title for the entirety of the first three and a half years, Saha says.
Aurora got closer to fulfilling its original goal earlier this month, too, when Amazon engineers very publicly turned off the last Oracle database powering Amazon Fulfillment, one of the company's largest units.
“We always wanted to migrate off Oracle database onto something home-grown,” Saha says. “That's one of the reasons we built Dynamo.”
Because Aurora is the fastest-growing product at AWS, because Oracle has clearly taken notice, because it's caused some kind of clash between Jassy and Ellison, and because it's running much of Amazon's business, it is also “something that Jeff [Bezos] pays a lot of attention to,” Saha said.
And all this success and attention, he says, “feels great.”