Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Improving Storage System Performance By Making the File System Aware of Background Activities on the Storage controller

IP.com Disclosure Number: IPCOM000011416D
Original Publication Date: 2003-Feb-19
Included in the Prior Art Database: 2003-Feb-19
Document File: 1 page(s) / 41K

Publishing Venue

IBM

Abstract

Today's storage control units often perform background activities (e.g., background copy for a point-in-time copy) which lead to temporary increased load on certain ranks(sometimes called RAID arrays). If the storage applications (e.g., SAN File systems) sitting on top of these control units were aware of this background activity, they could potentially redirect some of their requests to other ranks, thus improving overall system performance. In this invention, the storage controller notifies the storage application that some portion of the data, e.g., a RAID rank, it controls will deliver degraded performance. This information can be provided via an SNMP alert or other communication mechanism. The storage application then attempts to direct its activities away from this portion of the data. Several concrete means of doing this include: 1) allocating new storage from different portions of the storage controlled by this controller 2) using a mirroring approach, to direct the I/O to different portions of the storage 3) closing data base logs stored on the portion of the storage with degraded performance and opening new logs on different portions of the storage Other means exists.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 63% of the total text.

Page 1 of 1

  Improving Storage System Performance By Making the File System Aware of Background Activities on the Storage controller

We start from the observation that as disk capacities grow faster than transfer rate, RAID arrays will be spending an increasing percentage of their time in rebuild; we assume that the data availability issue will be solved. However, RAID drive rebuild, rather than the exception, should be thought of as the rule.

    Overall system performance can be improved if the system sitting on top of the storage controller does not use the degraded rank during the rebuild. Also, rebuild takes longer if, in addition to the rebuild-driven read/write activity, other read/write activities are served. Therefore, in addition, by avoiding using the rank during this period of degradation, we can have a faster rebuild and reduce the window in which data loss is possible due to an additional failure

    If the storage controller gives a notification (e.g., an SNMP alert), to the file system, or other code running on top of it, when a rebuild event is occurring on a particular rank, the higher level code could use this information to redirect I/O away from the rebuilding rank, e.g., to avoid where possible allocating from storage pools on this rank. Mirroring would be another mechanism which could be used to direct I/O to other ranks. As a third example a "database log" on the failed rank could be closed, and a new one started on a healthy rank.

More generally, today's stor...