12.07.2015 Views

Strategies for Data Protection - Brocade

Strategies for Data Protection - Brocade

Strategies for Data Protection - Brocade

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Highly Available Storagecharacteristic of RAID 1. The inverse of RAID 0+1 is RAID 10, in whichcase the mirroring occurs first as a virtual disk be<strong>for</strong>e striping isexecuted.Other RAID techniques attempt to integrate the per<strong>for</strong>mance advantageof data striping with alternative means to reconstruct data in theevent of disk failure. The most commonly deployed is RAID 5, whichstripes data across a disk set and uses block parity instead of mirroringto rebuild data. As data blocks are striped across multiple disks, aparity block is calculated using an eXclusive OR (XOR) algorithm andwritten to disk. If a disk fails, the data can be reconstructed on a newdisk from the parity blocks. In RAID 4, the parity blocks are written to asingle dedicated disk. This creates some vulnerability if the parity diskitself fails and incurs a write penalty, since every write must be parityprocessed on a single drive. RAID 5 reduces the write penalty by placingthe parity in<strong>for</strong>mation across multiple disks in the RAID set. As theparity data is generated, the array controller does not have to wait <strong>for</strong>the availability of a dedicated disk. As shown in Figure 10, RAID 5arrays typically house spare disks that can automatically be broughtonline and reconstructed in the event of disk failure. In this example, ifthe third disk in the set fails, the parity block on the fifth disk (P abcd)can be used to recreate both block C and the parity block (P efgh) <strong>for</strong>blocks E, F, G, and H.Figure 10. RAID 5 with distributed parity blocksThe primary benefit of RAID 5 is its ability to protect block data whileminimizing the number of disks required to guard against failure. Onthe other hand, the write penalty generated by parity calculation needshardware acceleration to improve per<strong>for</strong>mance and avoid an adverseimpact to upper-layer applications.<strong>Strategies</strong> <strong>for</strong> <strong>Data</strong> <strong>Protection</strong> 19

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!