Brandon Myers rocked his generals exam today. Brandon (perhaps atypically, but appropriate in this case) gave a literature-review style talk in which he focused on work in many fields related to the performance of generated code and hand-coded solutions for distributed query processing.

In-memory data management (IMDM) systems have achieved query processing performance that nears hand-coded implementations. This paper classifies work on how this is achieved. Availability of fast IMDM systems is motivation for incorporating them into more applications. We survey techniques for co-operation of application code with declarative queries. We find there remain challenges in the integration of high-performance parallel programming with IMDM. We present opportunities for new work in this area, and we describe our work on increasing the optimization opportunities of queries using a parallel language.

I left the room with a better understanding of the area, a growing appreciation for Brandon’s work, and confidence in his Ph.D. success. Read more about his work in other posts.

I went home with a migraine and feel asleep early, then ruined my night by waking up at 3am and penning an annoyed blog post about the latest horrible, pathetic excuse for “data in the news” to catch my eye.