Performance Engineering for In-Memory Databases: Models, Experiments and Optimization

The recent growth of interest for in-memory databases poses the question on whether established performance engineering methods such as analytical models, response surfaces and queueing simulation are effective in describing these database systems. In this talk, I will discuss our recent work on analytical models for performance assessment and optimization of inmemory databases. These include novel response time approximations under online analytical processing workloads to model thread-level fork-join and per-class memory occupation. I will then discuss the relative merits of performance modelling compared to experimental design methods that generate response surfaces and our recent experience on optimal workload placement in such systems.