Carnegie Mellon University Technical Report CMU-CS-03-124, April 2003. Superceded by VLDB 03, Berlin, Germany, Sept 9-12, 2003.
Jiri Schindler, Anastassia Ailamaki*, Gregory R. Ganger
Electrical and Computer Engineering
School of Computer Science
Carnegie Mellon University
Pittsburgh, PA 15213
Database systems work hard to tune I/O performance, but they do not always achieve the full performance potential of modern disk drives. Their abstracted view of storage components hides useful device-specific characteristics, such as disk track boundaries and advanced built-in firmware algorithms. This paper presents a new storage manager architecture, called Lachesis, that exploits a few observable device-specific characteristics to achieve more robust performance. Most notably, it enables efficiency nearly equivalent to sequential streaming even in the presence of competing I/O traffic. With automatic adaptation to device characteristics, Lachesis simplifies manual configuration and restores optimizer assumptions about the relative costs of different access patterns expressed in query plans. Based on experiments with both IBM DB2 and an implementation inside the Shore storage manager, Lachesis improves performance of TPC-H queries on average by 10% when running on dedicated hardware. More importantly, it speeds up TPC-H by up to 3X when running concurrently with an OLTP workload, which is simultaneously improved by 7%.
KEYWORDS: Database storage management, Performance evaluation