Asymmetric storage data distribution

Patent metadata

Field	Value
Publication number	US-10545683-B2
Application number	US-201615175065-A
Country	US
Kind code	B2
Filing date	Jun 7, 2016
Priority date	Feb 8, 2016
Publication date	Jan 28, 2020
Grant date	Jan 28, 2020

Abstract

Aspects include defining a first percentage of storage areas in an array of multiple persistent storage elements as hot storage areas and a second percentage of storage areas as spare storage areas such that remaining storage areas define a third percentage as cold storage areas. Each of the storage areas are assigned to either the hot group, the spare group or the cold group, respectively. A hot and cold storage area each include a first storage block on two different storage elements, and the hot storage area and the cold storage area each include a corresponding second storage block on a storage element different to the storage element on which the first respective storage block is stored. The storage blocks are distributed across the storage elements such that blocks of storage areas with the highest write rate of all storage areas are placed on a hottest storage element.

First claim

What is claimed is:

1. A method for operating an array of multiple persistent storage devices that includes a General Parallel File System (GPFS) and includes data stored in storage areas that are distributed across all of the storage devices of the array, each of the storage devices having a plurality of storage stripes, wherein each of the plurality of storage stripes comprises a storage block and wherein each stripe inhabits each device of the array, said method comprising:
receiving, from a system administrator of the system, a first percentage of said storage areas to be defined as hot storage areas and a second percentage of said storage areas to be defined as spare storage areas;
defining the remaining storage areas of said storage areas as cold storage areas such that a hot group of hot storage areas, a spare group of spare storage areas and a cold group of cold storage areas are built, wherein the spare storage area is not used during normal operation of the array of multiple persistent storage devices;
assigning each of said storage areas to either said hot group, said spare group or said cold group, respectively, wherein a hot storage area and a cold storage area each comprise a first storage stripe on a first plurality of storage devices, and wherein said hot storage area and said cold storage area each comprise a corresponding second storage stripe on a second plurality of storage devices different to said first plurality of storage devices on which said first respective storage stripe is stored;
distributing a plurality of storage stripes across all of the storage devices of the array such that stripes of storage areas with the highest write rate of all storage areas are placed on a hottest storage device;
defining at least one logical unit comprising a plurality of storage areas;
providing access to said at least one logical unit to store data;
tracking a total number of write accesses to at least one of each of said storage areas and each of said storage devices during operation of the array of multiple persistent storage devices; and
rebalancing write amplifications within the array of multiple persistent storage devices based on the tracked total number of write accesses, in response to writing to at least one of said storage devices during operation of the array of multiple persistent storage devices, wherein the rebalanced write amplifications are steered to at least one storage device of the array of multiple persistent storage devices determined to fail before other storage devices of the array of multiple persistent storage devices based on the tracked total number of write accesses.

2. The method according to claim 1 , wherein said storage areas are organized as a redundant array of independent disks.

3. The method according to claim 1 , wherein said storage stripes are all of equal size.

4. The method according to claim 1 , further comprising:
upon determining that said total number of write accesses to a storage device exceeds a predefined threshold value, moving said stripes being stored on said storage device to stripes of said spare storage areas.

5. The method according to claim 1 , further comprising:
prior to moving said one stripe of that storage area having a higher write rate than any of said stripes of storage areas located on said hottest storage device to said hottest storage device, moving a stripe of a storage area having a lowest write access rate of stripes on said hottest storage device to one of said spare storage areas.

6. The method according to claim 5 , further comprising:
marking said one stripe that has been moved to said hottest storage device as belonging to said spare group.

7. The method according to claim 1 , wherein each storage device in the array of multiple persistent storage devices has a write dependent wear level.

8. The method according to claim 1 , wherein a number of stripes in the spare storage area is at least equal to a number of storage stripes on a hottest storage device.

9. The method according to claim 1 , wherein write amplifications occur when a majority of storage stripes of a persistent storage device contain data.

10. The method according to claim 1 , further comprising replacing storage devices that have reached or exceeded an associated life-time based on the tracked total number of write accesses.

11. The method according to claim 1 , further comprising providing a notification to replace storage device determined to fail before an actual failure of the storage device occurs.

Assignees

• IBM

Inventors

• Kraemer Marco
• Otte Carsten
• Raisch Christoph

Classifications

• G06F3/06

  Digital input from, or digital output to, record carriers {, e.g. RAID, emulated record carriers or networked record carriers} · CPC title

• G06F3/0614

  Improving the reliability of storage systems · CPC title

• G06F3/061

  Improving I/O performance · CPC title

• G06F3/0649 Primary

  Lifecycle management · CPC title

• G06F3/0689

  Disk arrays, e.g. RAID, JBOD · CPC title