Browse Prior Art Database

End-to-End Performance Monitor and Bottleneck Analysis Algorithm with Application to IBM Storage Tank

IP.com Disclosure Number: IPCOM000031328D
Original Publication Date: 2004-Sep-21
Included in the Prior Art Database: 2004-Sep-21
Document File: 4 page(s) / 102K

Publishing Venue

IBM

Abstract

This article presents an algorithm to automatically drill down and isolate poorly performing logical volumes in the attached back-end storage devices in the IBM Storage Tank* architecture.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 43% of the total text.

Page 1 of 4

End-to-End Performance Monitor and Bottleneck Analysis Algorithm with Application to IBM Storage Tank

    This article presents a method for end-to-end performance monitoring and problem determination in the context of IBM Storage Tank* (ST) [1]. A novel algorithm is proposed to automatically drill down and isolate poorly performing logical volumes in the attached back-end storage devices in the ST architecture. This is accomplished by statistically correlating application level performance to that of the attached storage devices. Because the algorithm utilizes historical data, trend analysis can be performed, and all computations can be performed off-line. While the proposed method is presented in the context of ST, it has application in basic storage area network (SAN) configurations without ST as well. In the proposed method, performance monitoring is accomplished in three main steps. Firstly, obtain application level performance information from Application Response Measurement (ARM) [2] instrumented applications using agents and return this information to a proposed Performance Monitor (PM) server on a scheduled basis. At regular intervals the PM server stores this information to a dedicated database. Secondly, on a scheduled basis, obtain configuration and file mapping information by querying the ST metadata server, using custom PM agents, and return this information to the PM server. Once again this information is written to the PM database. Thirdly, combine and correlate all PM database performance information with that of attached storage performance information. The attached storage performance information is obtained from IBM Multiple Device Manager* (MDM), and focuses on

Figure: End-to-End Performance Monitor and Bottleneck Analyzer for Storage Tank

SAN logical units (LUNs). At this point, the proposed bottleneck analysis algorithm can be run to automatically isolate the poorly performing logical volumes (and a mapping of the attached hardware) to allow the SAN administrator to quickly pinpoint the source of performance problems. A diagram of the PM incorporated into the Storage Tank architecture is given in the

1

[This page contains 1 picture or other non-text object]

Page 2 of 4

Figure above.

    While this article presents a method for end-to-end performance monitoring and bottleneck analysis in the context of ST, the method can be applied in basic SAN configurations. The application level information would still come from ARM agents and be sent back to the PM server. The storage-level statistics would still come from MDM and be correlated with the application information. The only change would be within the mapping element. In ST, the file mapping information is obtained by querying the ST metadata server via command line interface. As an alternative, this mapping information can be obtained in basic SAN configurations without ST via the standard host bus adapter (HBA) API, namely by utilizing the HBA_GetFcpTargetMapping A...