Systems Seminar - CSE
The journey from faster to better
Add to Google Calendar
RocksDB has been adapted to a wide range of workloads. Early adopters needed a DBMS for low-latency & high-throughput workloads. Using RocksDB as an embedded DBMS eliminates network latency per request. When using RocksDB with a fast SSD we are able to get more IO throughput compared to other open-source DBMS that we use in production.
We continue to get new uses and new requirements for RocksDB. Faster average performance isn’t sufficient for OLTP, we also need minimal variance in response time. We will describe our efforts to improve the quality of service in RocksDB and plans to be efficient on a variety of storage devices including SSD, disk and hybrid (disk + SSD). We expect some of our open issues to be interesting to systems researchers. We will also share stories about fun times in the land of web-scale DBMS deployments.
Mark Callaghan worked with great teams to make MySQL better for web-scale deployments at Facebook & Google for 9+ years. His current focus at Facebook is database algorithms and storage systems for small data (OLTP) workloads. He also works with WebScaleSQL and RocksDB to make MySQL and MongoDB better. Prior to his web-scale work he spent many years working on RDBMS internals at Oracle and Informix. He invented and implemented a very fast general purpose sort algorithm for the Oracle RDBMS. He has an MS in CS from UW-Madison.