Azul Systems: Consistent Cassandra Performance with Zing
Read/Download White Paper (PDF)
Apache Cassandra is a NoSQL distributed database written in Java and designed for reliability and high scalability. Data nodes are spread across lots of commodity servers and even between datacenters. Information is kept in-memory for fast read and write response times, then “flushed” to disk when memory fills.
Because Cassandra is built in Java, it is subject to the limitations and performance hiccups associated with traditional Java Virtual Machines (JVMs). Common issues include response time outliers, frequent Java garbage collections, out of memory errors and system hangs. Companies deploying Cassandra spend extensive time tuning and tweaking the JVM to avoid these issues, then need to re-tune each time load (or the application accessing Cassandra) changes.
The Azul Zing runtime allows Cassandra to run smoothly, free of Java-related performance glitches and hangs. Zing is an innovative JVM that removes garbage collection as an issue. Now your deployment can have consistent response times all the time, without GC pauses, node hangs or out of memory errors, regardless of load or object allocation rates.
Zing also provides several runtime optimizations in addition to the pauseless garbage collector. These runtime optimizations enable the application to continue to perform where traditional JVMs would not. For example, Zing has some unique features that enable code to warm up sooner and remain consistent longer.