Affiliation:
1. Intel Labs Pittsburgh
2. École Polytechnique Fédérale de Lausanne
3. Carnegie Mellon University
Abstract
TPC-E is a new OLTP benchmark recently approved by the Transaction Processing Performance Council (TPC). In this paper, we compare TPC-E with the familiar TPCC benchmark in order to understand the behavior of the new TPC-E benchmark. In particular, we compare the I/O access patterns of the two benchmarks by analyzing two OLTP disk traces. We find that (i) TPC-E is more read intensive with a 9.7:1 I/O read to write ratio, while TPC-C sees a 1.9:1 read-to-write ratio; and (ii) although TPC-E uses pseudo-realistic data, TPC-E's I/O access pattern is as random as TPC-C. The latter suggests that like TPC-C, TPC-E can benefit from SSDs, which have superior random I/O support. To verify this, we replay both disk traces on an Intel X25-E SSD and see dramatic improvements for both TPC-C and TPC-E.
Publisher
Association for Computing Machinery (ACM)
Subject
Information Systems,Software
Cited by
53 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献