Big data technologies enable the processing of massive data sets and promise to deliver new and innovative solutions. But consistent, low latency performance isn’t a guarantee for all Java-based big data applications. Careful attention to product choices, runtime components and deployment topologies are essential to maximizing the values of these new big data solutions, including Apache Cassandra and supporting components such as Spark and Solr.
A benchmark study was done comparing the response time performance of two different JVMs, Azul Zing and Oracle HotSpot, while running Apache Cassandra at different throughput levels. Benchmark results were derived using the Cassandra-stress framework, a utility for load testing and benchmarking a Cassandra cluster, and the jHiccup JVM measurement tool.
The testing methodology was designed to measure JVM response time consistency based on percentiles and Cassandra application max outliers at different throughput rates. The benchmark was configured with three Cassandra data nodes and one server. The jHiccup measurement tool was used to capture and graph the response time distribution of the two JVMs for all benchmark runs.
Download the benchmark study to find out about the performance results.