View article

Over the last decade, we have seen a rise in demand for analytical data processing. This trend is driven by an increase in volume and velocity of data that is being created and by companies seeking to unlock its potential. Examples include vehicle telemetry, health, and industrial data. At the same time, advances in hardware and machine learning enable system builders to create ever faster and smarter database systems.

This thesis makes three contributions to the design and implementation of such systems. First, we compare mainmemory databases with modern streaming systems using a telecommunications benchmark, identify performance and usability gaps, and explore extensions to database systems. These extensions include user-space networking for faster client-server communication and a scale-out architecture. Second, we propose an approach to processing geospatial point data in memory. In …