Inside Aurora: how disruptive is Amazon’s MySQL clone?

Why MySQL?

How will MySQL compatibility help Amazon take market share from proprietary database managers?

“Lots of enterprises are using MySQL,” claims Jassy. “They would like to move more workloads to MySQL but it is challenging to get the performance that they want. We wanted something that customers had some comfort and experience with, and that didn’t make them feel locked in.”

The snag is that porting applications from proprietary database engines tends to be difficult because of vendor-specific features. The low-hanging fruit is applications that already use MySQL, and if that goes well, perhaps businesses will consider Aurora for new work.

Even among open source relational database managers, many developers prefer PostgreSQL. Might Amazon deliver an Aurora-like clone of that as well? “The sort of things we have changed are not specific to MySQL,” says Gupta, implying that the work of separating the logging and storage components could equally be done for other database engines. At re:Invent, PostgreSQL compatibility was a common request.

The open source question

Amazon uses a lot of open source technology, but how much does it give back? Asked about this at re:Invent, Jassy replied tersely. “We contribute to the open source ecosystem for many years. Zen, MySQL space, Linux space, we’re very active contributors, and will continue to do so in future.”

Amazon's Andy Jassy, Senior VP of AWS

That raises the question: since MySQL is open source, has Amazon used any of that code in its own Aurora implementation? I asked marketing executive Paul Duffy, who said:

“That’s why we say it is compatible with MySQL. If you run the MySQL compatibility tool that will all check out. We don’t disclose anything about the inner workings of the service.”

That was hardly a complete answer. A similar question directed at Gupta drew this response:

“All of the storage is created within Amazon. Of course the database started from MySQL so there’s some dependencies there – there’s an awful lot of that database that has changed.”

Power to disrupt?

Aurora is in limited preview, so it is too early to judge its capabilities. The theory behind it is that a database engine designed to run on a cloud-scale storage layer can work differently and in some ways better than one designed to work with conventional storage, and that does make sense. Amazon has a pure cloud play which gives it more freedom versus other vendors who deliver on-premise versions of their database products, and its commodity pricing should make the service attractive.

The downside of Aurora is that it is a MySQL clone, and MySQL is not the first choice for most enterprises. In addition, database administrators are rightly a conservative bunch, and will want to see Amazons claims for resilience tested in the real world before trusting it with critical data. ®