Amazon Web Services Inc. (AWS), has made it generally available an in-memory cache that increases the performance of Amazon DynamoDB, its NoSQL database service.
Amazon DynamoDB Accelerator (DAX), which was announced in April preview, promises a 10-fold performance improvement in DynamoDB queries.
AWS states that DAX does all of the heavy lifting to add in-memory speed to DynamoDB tables without you having to manage cache invalidation or data population. “Now you can focus on building great apps for your customers and not worrying about performance at scale.”
Yesterday’s blog post by Jeff Barr, AWS spokesperson, stated that the company used the preview period for fine-tuning the service. The service’s reach has been extended to additional AWS Regions, to increase the total to five.
The number of SDKs that developers can use DAX has not been expanded. Barr stated that Java’s DAX SDK is the only one currently available. However, AWS is working on other SDKs.
“DAX returns cached responses in microseconds, making it a great fit for eventually-consistent read-intensive workloads,” Barr said. “DAX supports DynamoDB API and is seamless and simple to use. You can simply create your DAX cluster as a managed service and use it to target your existing reads or writes. You don’t need to worry about cluster maintenance, patching, replication, fault management, or cluster maintenance.
Werner Vogels, the company’s chief technology officer, wrote a detailed post yesterday on his blog “All Things Distributed”, giving a glimpse into the history and technical details of DAX. He said that developers only need to point an application to a DAX Endpoint with three lines of Java code.
Vogels stated that DAX is a fully managed caching system that is API-compatible to DynamoDB. This means that developers don’t need to rewrite DynamoDB applications to use DAX. Instead, you can point your existing Java application to a DAX endpoint using the DAXSDK for Java. DAX takes care of the rest.
As referenced on the DynamoDB pricing page, users pay for the DAX add-on service on a per-node-hour-consumed model, starting at $0.269 per hour in the US East (Northern Virginia) and US West (Oregon) regions — other regions are slightly more.
You can find more information in the DAX documentation.