Matthew Pope

Software Engineer at Datometry

Matthew Pope is a Software Engineer working with Erlang in the database virtualization field. He has previous experience working in the video game industry, working on scaling Ejabberd as well as real-time match making. He wants to see the BEAM ecosystem thrive as more advanced tools and languages are created.

High throughput data conversion for database virtualization

Audience: Intermediate

Hyper-Q is a database virtualization application written in Erlang. It emulates the syntax and semantics of a guest database over a host database, which involves cross-compiling SQL from one dialect to another. Additionally, it emulates functionalities that are not offered by the host database. Tuning Hyper-Q for new SQL workloads involves identifying bottlenecks in the cross-compilation process. This presentation will provide an overview of the diagnosis process for identifying bottlenecks through profiling and will also focus on some architectural changes made to address them.

The main objective of this initiative was to reduce memory consumption during processing, thereby improving the throughput of queries that can be processed through Hyper-Q. This, in turn, assists customers in transitioning from legacy data warehouses to cloud databases with minimum need for application rewrites saving time and effort.

OBJECTIVES: This talk will demonstrate the use of profiling tools, diagnostic process and provide an overview of system architecture at Datometry.

AUDIENCE: Distributed systems developers, architects, Database internals and big data processing enthusiasts.