What we did
Pythian improved the performance and stability of Cassandra through anti-entropy repairs, drastically improved internal latencies, and increased cluster availability using a DataStax optimization report.
- Optimized the Cassandra database cluster performance. When the SoundCloud team implemented Cassandra Reaper, they faced several issues making the tool operational. The Pythian team stepped in, and now Cassandra anti-entropy repairs without any issues.
- Enabled internal speed by improving tail read latencies. Pythian improved the read latencies on the 99th percentile from 250ms to 40ms. This allowed the SoundCloud team to provide internal customers the support they needed in a timely fashion.
- Changed a Cassandra cluster from a single datacenter to a multi-datacenter topology. Pythian increased data availability for a Cassandra cluster by helping provision a new datacenter. They did this while adhering to strict topological constraints.
- Helped backend teams configure consistency levels. Pythian also investigated the consistency levels in Cassandra. Pythian created a report explaining how the DataStax Java driver for Cassandra could be optimized while using the multi-cluster.