"With MySQL you might be doing a nightly snapshot [of data]. Now, with Aurora, it is happening all the time, automatically," Tavis said.
In terms of resilience, Aurora can detect a database failure and recover in less than a minute, without the need to rebuild (or "warm") the database caches that are needed to speed response times. In the case of a permanent failure, Aurora can automatically failover to a replica without losing any data.
The company also designed Aurora to scale easily to large workloads: Each database instance can hold up to 64 terabytes of data.
This is not Amazon's first relational database service. RDS already serves as a backend for MySQL, Oracle, Microsoft SQL Server, and PostgreSQL databases, automating capabilities such as disk allocation, point-in-time recovery and performing snapshots of the data.
Aurora adds some additional capabilities on top of RDS, in terms of automating backups, scaling and pre-setting the tuning for high scalability usage, Tavis said. Amazon engineers found ways to minimize delays created by database process threads and lock contentions. In addition, Aurora is designed to run on speedy solid-state disks (SSDs).
Amazon is also pitching Aurora as a way for enterprises to escape the lock-in from commercial database vendors, such as Oracle (which now manages the open source MySQL, inherited from Oracle's 2010 purchase of Sun Microsystems). When the service goes live, it will cost US$0.29 per hour for each large instance, with no upfront costs.
With traditional enterprise database vendors, "there is a high amount of lock-in and they have punitive licensing terms--not just allowing very little flexibility in moving to the cloud, but also in the auditing and fining of their customers," Jassy said during a Wednesday keynote at which Aurora was announced. "That is why you see so many enterprises and companies trying to figure out how to move as many workloads as possible to the more customer-friendly open source database engines."
Sign up for Computerworld eNewsletters.