Considerations for RAID-6 Availability and Format/Rebuild Performance on the DS5000
An IBM Redpaper publication
Published 03 February 2009, updated 30 March 2010
IBM Form #: REDP-4484-00
Rate and comment
Authors: Alex Osuna, Siebo Friesenborg, Michael D. Roll
This IBM® Redpaper publication documents considerations for RAID-6 availability and format/rebuild performance on the DS5000. It was created by Siebo Friesenborg of Storage ATS for Americas Group using data provided by Michael Roll of Modular Systems Development. The DS5000 implementation of RAID-6 was used to generate materials in this paper. Other implementations may be similar (and they may not). No attempt was made to cover a breadth of products.
Topics covered in this paper include:
* Basic information about how the DS5000 handles write operations. The first part of this subject deals with the handling of write requests in cache write storage. It is the same regardless of the RAID level or the type of physical/logical drive.
* When a certain amount of cache write storage is used, programming is initiated that frees up cache write storage by writing data to the high density drives (HDDs). The number of operations required to physically write data to disk depends on the RAID used. This is explained for RAID-5, RAID-6, and RAID-10.
The next section of this paper discusses a method of estimating what the additional load on HDDs would be if RAID-6 were used instead of the current RAID. It is based on actual data gathered from a production DS3000, DS4000® , or DS5000 through the Performance Monitor.
A third topic is measurements taken of the time to initially format arrays and the time to recover from a failed disk. For these measurements we used 300 Gigabyte, 15 K RPM, FC drive. No servers were used for this. Rather, selecting an HDD on the GUI and using the fail drive button of the Advanced Recovery window was used to cause rebuild.
Table of contents
Follow IBM Redbooks
Follow IBM Redbooks