Enhancing disk array performance via variable parity based load balancing

• US 20020124137A1
• Filed: 01/29/2002
• Published: 09/05/2002
• Est. Priority Date: 01/29/2001
• Status: Abandoned Application

• Current Assignee: Braodband Storage LLC, Snap Appliance, Inc. (Silicon Valley Technology Partners LLC)
• Original Assignee: Adaptec Incorporated (Steel Partners Holdings LP)
• Inventors: Priester, Wilbor George, Schweitzer, James R., Bolstad, Gregory D., Staub, John R., Randall, Jay G., Ulrich, Thomas R.
• Application Number:
  US10/060,916
• Publication Number:
  US 20020124137A1
• Time in Patent Office: Days
• US Class Current: 711/113
• CPC Class Codes: G06F 11/1084 Degraded mode, e.g. caused by single or multiple storage removals or disk failures
• CPC Class Codes: G06F 2206/1012 Load balancing
• CPC Class Codes: G06F 2211/1004 Adaptive RAID, i.e. RAID system adapts to changing circumstances, e.g. RAID1 becomes RAID5 as disks fill up
• CPC Class Codes: G06F 2211/1028 Distributed, i.e. distributed RAID systems with parity
• CPC Class Codes: G06F 3/0613 in relation to throughput
• CPC Class Codes: G06F 3/0614 Improving the reliability of storage systems
• CPC Class Codes: G06F 3/064 Management of blocks
• CPC Class Codes: G06F 3/0643 Management of files
• CPC Class Codes: G06F 3/0646 Horizontal data movement in storage systems, i.e. moving data in between storage devices or systems
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

A system and method for enhancing disk array performance via variable-parity-based load balancing is described. The system and method operate on a computer storage system that includes a plurality of disk drives for storing distributed parity groups. Each distributed parity group includes storage blocks. The storage blocks include one or more data blocks and a parity block that is associated with the one or more data blocks. Each of the storage blocks is stored on a separate disk drive such that no two storage blocks from a given parity set reside on the same disk drive. The computer storage system further includes file system metadata to describe a location of each of the storage blocks. The computer storage system further includes a load-balancing module to dynamically move storage blocks of the distributed parity groups among the plurality of disk drives to balance transaction loading of the plurality of disk drives and thereby to improve throughput of the computer storage system. The load-balancing module is configured to update the file system metadata when a storage block is moved.

Claims

1. A computer storage system, comprising:

a plurality of disk drives for storing distributed parity groups, each distributed parity group comprising storage blocks, said storage blocks comprising one or more data blocks and a parity block associated with said one or more data blocks, each of said storage blocks stored on a separate disk drive such that no two storage blocks from a given parity group reside on the same disk drive;

file system metadata to describe a location of each of said storage blocks; and

a load-balancing module to dynamically move storage blocks of said distributed parity groups among said plurality of disk drives to balance loading of said plurality of disk drives and thereby improve throughput of said computer storage system, said load-balancing module configured to update said file system metadata when a storage block is moved.

2. The computer storage system of claim 1, wherein a size of a first distributed parity group is larger than a size of a second distributed parity group.

3. The computer storage system of claim 1, further comprising metadata to specify which disk drive in said plurality of disk drives contains a desired storage block.

4. The computer storage system of claim 1, wherein said disk drives are provided to a Fibre Channel network.

5. The computer storage system of claim 1, wherein a file is comprised of one or more of said distributed parity groups.

6. The computer storage system of claim 1, wherein said file system metadata comprises information to specify a logical block address for each storage block in a distributed parity group.

7. The computer storage system of claim 1, wherein an extent of a first distributed parity group of a file is larger than an extent of a second distributed parity group of said file.

8. The computer storage system of claim 1, wherein said load-balancing module runs in a background process during file system operations.

9. A method for load balancing in a computer network, comprising:

determining a size of a parity group in response to a write request, said size describing a number of data blocks in said parity group;

arranging at least a portion of data from said write request according to said data blocks;

computing a parity block for said parity group;

storing each of said data blocks on a separate disk drive of a plurality of disk drives such that no two data blocks from said parity group reside on the same disk drive;

storing said parity block on a separate disk drive that does not contain any of said data blocks; and

moving said data blocks among said plurality of disk drives to reduce loading on one or more of said plurality of disk drives.

10. The method of claim 9, further comprising storing metadata to describe a disk and logical block location of each of said data blocks and said parity block.

11. The method of claim 9, wherein said moving comprises combining a first parity group having a first size and a second parity group having a second size to produce a combined parity group having a third size, wherein said third size specifies a number of blocks that is, at most, one less than the number of disk drives available to store data from said parity group.