Data recovery from multiple failed data blocks and storage units

• US 7,428,691 B2
• Filed: 11/12/2003
• Issued: 09/23/2008
• Est. Priority Date: 11/12/2003
• Status: Expired due to Fees

• Current Assignee: Norman Ken Ouchi
• Original Assignee: Norman Ken Ouchi
• Inventors: Ouchi, Norman Ken
• Primary Examiner(s): Lamarre; Guy J
• Application Number:
  US10/706,881
• Publication Number:
  US 20050114726A1
• Time in Patent Office: 1,777 Days
• Field of Search: 714/763, 714/773, 714769-771, 711/162
• US Class Current: 714/770
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems
• CPC Class Codes: G11B 20/1833 by adding special lists or symbols to the coded information G11B20/1806, G11B20/1866 take precedence

Abstract

In the past, storage unit (disk drive) failures were the primary cause of data loss in a storage system. With higher unit reliability and higher bit density, random bit errors have become the primary cause of data loss. Most data recovery mechanisms treat reconstruction of redundant information on the same level as data reconstruction. In reality, data reconstruction is more important and asymmetry between data protection and redundant information protection provides trade-offs of data recoverability against performance. The present invention provides data recovery from both a first number of data block failures due to random bit failures and a second number of storage unit failures while providing update write performance equivalent to data protection mechanisms with lower data recovery capabilities. The level protection from number of data block failures, the number of unit failures, and update write performance are parameterized to select a desired combination.

Claims

1. A storage system with data recovery from M failed blocks per stripe or J failed storage units comprising N, N>

0 data blocks stored on N storage units and a first error correction code that generates M, M>

0 redundant blocks from the N data blocks where the N data blocks and M redundant blocks form a stripe, wherein the M redundant blocks are stored on J, J less than or equal to M but at least one additional storage units such that K failed data blocks, K less than or equal to M blocks are regenerated from the remaining N+M−

K blocks of the stripe;

or P storage units fail, P less than or equal to J containing data blocks wherein the data blocks in the failed storage units are regenerated from the remaining data blocks and remaining redundant blocks of the stripe.

2. The storage system of claim 1 wherein the storage for the redundant blocks is rotated among the N+J storage units such that the storage requirement is evenly distributed.

3. The storage system of claim 1 wherein the storage for the redundant blocks is rotated among the N+J storage units such that the storage accesses to the redundant blocks are uniformly distributed.

4. The storage system of claim 1 wherein the storage system provides additional data recovery from J failed storage units and L failed blocks per stripe wherein:

L, L less than or equal to M redundant blocks that are copies of the M redundant blocks of a stripe are stored on the storage unit with the most recent data block update for the stripe; and

in the event of failure of storage units with the M redundant blocks, the L copies of the redundant blocks are used to reconstruct up to L failed blocks of the stripe.

5. The storage system of claim 1 wherein the storage system provides additional data recovery from J failed storage units and L failed data blocks per stripe wherein:

L, L less than or equal to M redundant blocks that are copies of the M redundant blocks of a stripe are stored on the storage unit with the most recent data block update for the stripe; and

the storage blocks for the L copies of the redundant blocks are assigned as needed from a cool of storage blocks.

6. The storage system of claim 1 wherein the storage system provides additional data recovery from J failed storage units and L failed blocks in the stripe and R failed blocks for each stripe of a second stripe structure within each storage unit wherein:

L, L less than or equal to M redundant blocks are copies of the M redundant blocks of a stripe and are stored on the storage unit with the most recent data block update for the stripe; and

grouping S, S>

0 blocks stored on a storage unit including one block from the stripe, a second error correction code generates R redundant blocks from the S blocks and the R redundant blocks are stored on the storage unit to form a second stripe;

such that V failed data blocks, V less than or equal to R are regenerated from the remaining S+R−

V blocks of the second stripe.

7. The storage system of claim 1 wherein the storage system provides additional data recovery from J failed storage units and R failed blocks for each stripe of a second stripe structure in each functioning storage unit where:

for S, S>

0 blocks stored on a storage unit including one block from the stripe, a second error correction code generates R redundant blocks from the S blocks; and

the R redundant blocks are stored on the storage unit to form a second stripe;

such that V failed data blocks, V less than or equal to R blocks are regenerated from the remaining S+R−

V blocks of the second stripe.

8. A storage system with data recovery from L failed blocks per stripe comprising N, N>

0 data blocks stored on H, H greater than or equal to N+1 storage units, one data block per storage unit, and a first error correction code that generates M, M>

0 redundant blocks from the N data wherein;

the N data blocks and the M redundant blocks form a stripe such that K failed data blocks K less than or equal to M blocks are regenerated from the remaining N+M−

K blocks of the stripe; and

L, L less than or equal to M redundant blocks are stored on the storage unit with the most recent data block update;

such that when one or more of the M redundant blocks are not accessible, T, T less than or equal to L data blocks are regenerated from the remaining N+L−

T blocks of the stripe.

9. The storage system of claim 8 wherein the storage blocks for the L redundant blocks are assigned as needed from a pool of storage blocks.

10. The storage system of claim 8 wherein the storage system provides additional data recovery from J failed storage units or M failed blocks per stripe wherein:

the number of data blocks, N, equals the number of storage units, H, each with a data block from the stripe; and

the M redundant blocks for the stripe are stored on J, J less than or equal to M but at least one additional storage units;

such that P storage units fail, P less than or equal to J containing data blocks wherein the data blocks on the failed storage units are regenerated from the remaining data blocks and redundant blocks of the stripe.

11. The storage system of claim 8 wherein the storage system provides additional data recovery from J failed storage units or M failed blocks per stripe wherein the number of data blocks, N, equals the number of storage units, H, each storing a data block from the stripe, and the M redundant blocks for the stripe are stored on J, J less than or equal to M but at least one additional storage units and the storage requirement for the M redundant blocks is rotated among the H+J storage units so the storage requirement is equally distributed.

12. The storage system of claim 8 wherein the storage system provides additional data recovery from J failed storage units or M failed blocks per stripe wherein the number of data blocks, N, equals the number of storage units, H, each storing a data block from the stripe, and the M redundant blocks for the stripe are stored on J, J less than or equal to M but at least one additional storage units and the storage requirement for the M redundant blocks is rotated among the H+J storage units so the storage accesses to the redundant blocks are uniform.

13. The storage system of claim 8 wherein the storage system provides additional data recovery from J failed storage units and L failed blocks per stripe and R failed blocks per second stripe within a storage unit or M failed blocks per stripe and R failed blocks per second stripe within a storage unit wherein:

the number of data blocks, N, equals the number of storage units, H, each storing a data block from the stripe;

the M redundant blocks for the stripe are stored on J, J less than or equal to M but at least one additional storage units and for S, S>

0 blocks stored on a storage unit including one block from the stripe, a second error correction code generates R redundant blocks from the S blocks and the R redundant blocks are stored on that storage unit;

such that V failed data blocks, V less than or equal to R are regenerated from the remaining S+R−

V blocks of the second stripe.

14. A storage system with data recovery from R failed blocks per second stripe within a storage unit and J failed storage units or M failed blocks per first stripe across storage units and R failed blocks per second stripe within a storage unit comprising:

N, N>

0 data blocks stored on N storage units; and

a first error correction code that generates M, M>

0 redundant blocks from the N data blocks wherein the N data blocks and M redundant blocks form a first stripe across storage units such that K failed data blocks, K less than or equal to M are regenerated from the remaining N+M−

K blocks of the first stripe; and

the M redundant blocks are stored on J, J less than or equal to M but at least one additional storage units; and

S, S>

0 blocks stored on a storage unit including one block from the first stripe and a second error correction code that generates R, R>

0 blocks from the S data blocks wherein the S blocks and R redundant blocks form a second stripe stored within the storage unit such that V failed data blocks, V less than or equal to R blocks are regenerated from the remaining S+R−

V blocks of the second stripe;

such that the storage system can regenerate R blocks of failed data blocks for an S data blocks stripe within a storage unit and regenerate either data blocks for J failed storage units or M failed data blocks per N data block stripe across the storage units.

15. The storage system of claim 14 wherein the storage requirement for the M redundant blocks is rotated among the N+J storage units so that the storage requirement is evenly distributed.

16. The storage system of claim 14 wherein the storage requirement for the M redundant blocks is rotated among the N+J storage units so that the storage accesses to the redundant blocks are uniformly distributed.

17. The storage system of claim 14 wherein L, L less than or equal to M redundant blocks are L copies of the M redundant blocks are stored on the storage unit with the most recent data block update and the storage system provides additional data recovery from R failed blocks per second stripe and L failed blocks per first stripe across the storage units and J failed storage units.

18. The storage system of claim 14 wherein L, L less than or equal to M redundant blocks are L copies of the M redundant blocks are stored on the storage unit with the most recent data block update and the storage system provides additional data recovery from R failed blocks per second stripe and L failed blocks per first stripe across the storage units and J failed storage units wherein the storage blocks for the L redundant blocks are assigned on demand from a pool of storage blocks.