[DB Seminar] Spring 2017: Viktor Leis
Managing data sets that are larger than RAM has always been one of the most important tasks for database systems. Traditional systems cache fixed-size pages in an in-memory buffer pool that has complete knowledge of all page accesses and transparently loads/evicts pages from/to disk. While this approach is effective at minimizing the number of I/O operations, it is also one of the main reasons why disk-based systems are slow. For this reason, main-memory database systems abandon buffer management altogether, which makes handling data sets that are larger than main memory very difficult.
In this work, we design a novel storage manager that is optimized for modern hardware. Our evaluation, which is based on TPC-C and a number of micro benchmarks, shows that our approach has only 3% overhead in comparison with a pure in-memory system when all data resides in main memory. At the same time, like a traditional buffer manager, our storage manager is fully transparent and can manage very large data sets effectively. In contrast to traditional buffer managers, our storage manager is also highly scalable on multi-core CPUs.
Viktor Leis is a postdoctoral researcher at the Technical University of Munich. He does research on core database topics, including query processing, concurrency control, index structures, storage, and query optimization. Viktor's 2016 dissertation, which was done in the context of the main-memory database system HyPer, received the bi-annual dissertation award of the German-speaking database community (DBIS). He also received the best paper award at ICDE 2014 for his work on hardware transactional memory.