Configuring and Scheduling an Eager-Writing Disk Array for
a Transaction Processing Workload

Chi Zhang¹ - Xiang Yu¹ - Arvind Krishnamurthy² - Randolph Y. Wang¹

Abstract:

Transaction processing applications such as those exemplified by the TPC-C benchmark are among the most demanding I/O applications for conventional storage systems. Two complementary techniques exist to improve the performance of these systems. Eager-writing allows the free block that is closest to a disk head to be selected for servicing a write request, and mirroring allows the closest replica to be selected for servicing a read request. Applied individually, the effectiveness of each of these techniques is limited. An eager-writing disk array (EW-Array) combines these two complementary techniques. In such a system, eager-writing enables low-cost replica propagation so that the system can provide excellent performance for both read and write operations while maintaining a high degree of reliability. To fully realize the potential of an EW-Array, we must answer at least two key questions. First, since both eager-writing and mirroring rely on extra capacity to deliver performance improvements, how do we satisfy competing resource demands given a fixed amount of total disk space? Second, since eager-writing allows data to be dynamically located, how do we exploit this high degree of location independence in an intelligent disk scheduler? In this paper, we address these two key questions and compare the resulting EW-Array prototype performance against that of conventional approaches. The experimental results demonstrate that the eager-writing disk array is an effective approach to providing scalable performance for an important class of transaction processing applications.