News

Azul Launches New Service to Optimize Apache Cassandra

• By John K. Waters
• July 31, 2015

Java runtime maker Azul Systems Inc. has announced a new consulting service for enterprises facing Java-related "barriers" that affect their deployments of the Apache Cassandra database management system (DBMS). The new Cassandra HealthCheck (Cassandra HC) combines onsite and remotely delivered runtime analytic services, the company said, and leverages both open source and proprietary tools.

Cassandra is an open-source distributed DBMS originally developed by Facebook and built on Amazon's DynamoDB and Google's BigTable. It's designed to store and manage large amounts of structured data across many commodity servers. All nodes in the system are equal -- there's no master node -- which means that there's no single point of failure.

Cassandra was written in Java, which means that it's subject to the limitations of the Java runtime platform. "Many users do not understand the negative impact a Java runtime can have on Cassandra response time, reliability and throughput," said Scott Sellers, CEO and president of Azul Systems.

Garbage Collection (GC), for example (which Sellers has called "the Achilles heel of Java"), can affect performance, because the process freezes an application while memory is defragmented and compacted. The GC process causes random Cassandra pauses that lead to response time inconsistency, increased time to data consistency, and can even trigger serious conditions like cascading node failures, where multiple stalled Cassandra nodes trigger cluster-level outages and crashes, he said.

"We are seeing strong interest in the Apache Cassandra NoSQL database, particularly for applications that require a combination of low-latency performance and a highly available, distributed architecture," said Matt Aslett, research director in the Data Platforms and Analytics group at 451 Research, in a statement. "Garbage collection issues can hit any Java deployment, but are particularly significant where low-latency performance is a requirement. With Azul's new Cassandra HealthCheck offering, Cassandra users can leverage the company's JVM and performance tuning expertise, allowing for more consistent and reliable deployments while focusing their attention on the application itself rather than spending time trying to tune away JVM issues."

Sunnyvale, Calif.-based Azul's flagship product, Zing, is a Java Virtual Machine (JVM) based on Oracle's HotSpot JVM, a core component of Java SE. It's a "no-pause" JVM designed to eliminate GC pauses, a long-standing challenge for Java developers. This pauselessness, which Azul calls "generational pauseless garbage collection" (GPGC), enables Java app instances to scale dynamically and reliably.