Abstract
Discover how the enhanced performance and reliability of Amazon Aurora will help AWS customers reduce performance bottlenecks in their applications. The relatively low cost of Aurora will tempt many customers to migrate workloads to this implementation of RDS.
Sample
This white paper provides an overview of one of the latest offerings from cloud leader Amazon Web Services (AWS). It's a product that will enable powerful, massively scalable relational databases in Amazon's cloud environment. The product was designed, built, and tested in secret over the past three years, and now it is nearly ready for production workloads. Some of its detailed design remains confidential, so we'll present the released features and functionality, as well as pricing for this exciting new product.
Background
In November 2014, at Amazon Web Services' annual conference re:Invent, Amazon's CTO Werner Vogels introduced a series of revolutionary new products. Among those new products is the product described in this paper, which is based on AWS's hugely successful Relational Database Service (RDS). So what's this new product, you ask? It's Amazon Aurora for RDS. Aurora is MySQL compatible database solution that will enable highly scalable databases running across multiple AWS availability zones, at a very low price point.
What is Amazon Aurora?
Aurora is a MySQL compatible database engine that far exceeds the current size limitations of relational databases running on RDS. It spans three AWS availability zones within a region. The great news is that if you already use or are familiar with RDS, you're already going to be comfortable configuring Aurora. Aurora utilizes many of the same configuration options as other flavors of RDS. You simply have a new dimension to the implementation: greater scale and performance. Aurora will be a game changer in both the database and cloud markets.
An Introduction to Amazon RDS
For those readers who are unfamiliar with RDS, this section will give you a high level view of its features and functionality. (If you're familiar with RDS, please skip to the "Getting Started with Aurora" section.)
RDS provides a managed environment for customers to store their databases. At present there are four supported database options:
- Microsoft SQL
- Postgres SQL
- Oracle
- MySQL
The purpose of RDS is to provide a managed service for relational database use cases, and thus allow companies to offload the "heavy-lifting" of database management. In other words, AWS will manage the database admin tasks, while customers focus on their data.
RDS works similarly to Amazon Elastic Compute Cloud EC2. The customer chooses a virtual machine, or "instance" type for their database. Based on the use case, customers can choose instance types based on various levels of CPU, memory, and storage options. The customer can scale the instance type up or down as needed. Pricing is based on the instance type chosen and billed hourly. Aurora will be billed the similarly.
The database patching and backups are carried out by AWS on behalf of the customer. The backup retention period is customer configurable for backups (up to thirty-five days). In terms of point-in-time recovery, customers can choose to recover data up to five minutes in the past. In terms of patching, customers can choose whether to allow AWS to patch the database and when this should be done. The customer can then set this "maintenance window," and it's configurable via the management console and the API.
Although you can configure a single availability zone setup for RDS, AWS provides a multiple availability zone (multi-AZ) RDS option for high availability and disaster recovery purposes. In the multi-AZ configuration, RDS creates a primary and secondary database instance in two separate availability zones. A primary database instance is placed in one availability zone, and a secondary is placed in a second availability zone. In a third availability zone, a witness instance runs to observe the health of the primary instance. If that instance becomes unhealthy, the witness will cut over to the primary, switch the primary RDS endpoint address (or CNAME DNS record thereof) to the secondary instance. This means that there is very little downtime for the application using the database and there is no action required on the part of the customer to get the database back up and running.