Browse Prior Art Database

Publication Date: 2014-Jan-06
Document File: 4 page(s) / 74K

Publishing Venue

The Prior Art Database


This disclosure presents a method to improve the performance of array rebuilding. For most of the RAID Array implementation, the IO will have large impact to the performance of rebuilding process. In the worst cases with heavy workload, the duration of rebuilding maybe 10 times longer than normal case. This disclosure presents a method to redirect the IO during the array rebuilding process and then merge the IO back to the original array when the rebuilding is done. With such method, we'll get better and consistent rebuilding performance.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 01 of 4


In an SSD RAID, when one drive fails, the data will be rebuilt from other available drives to the spare drive, this process is called RAID rebuilding;

For SSD, what we've seen those issues:

- If there is NO IO against the being rebuild SSD RAID, it is fast to do the rebuild;

- But whenever there is any IO against the being rebuild SSD RAID, the time used to complete the rebuild will be much longer (eg. 3 times longer).

And the long rebuild period of an SSD RAID may cause many issues like:

- Normally, the will be performance loss due to array rebuilding; We've seen many cases that the customer's application are impacted due to RAID rebuild.

- The longer the array rebuilding lasts, the more often we'll have a chance to lose another drive during the rebuilding process and then may causes data loss.

There are already patents which are addressed to improve the RAID rebuild performance:

For example, the proposal in this patent is to migrate the hottest data from the being rebuild RAID to another spare array

But the drawbacks of this solution are:

- It is also impossible to detect when the RAID will start rebuilding and move the hottest data to the spare RAID;

So for normal case, the migration of hottest data happen in the process of RAID rebuild,

during the migration of the hottest data, there still be IO against the those data which will

then impact both the rebuild/application performance;

- There will be extra overhead to the rebuilding RAID introduced by the hottest migration which might have negative impact to the rebuild performance;

- The data placement will be changed due to RAID rebuild (as the hottest data are migrated due to another RAID);

In this invention, this solution is provided to improve the RAID rebuild performance and also IO performance during RAID rebuild:

- Introduce an Spare RAID into the system

- When one drive fails in one RAID and the RAID start rebuilding, all write IO against the rebuilding RAID is redirected to the Spare RAID;

- Keep track what block has been written to the Spare RAID

- When the rebuild of the failed RAID is complete, copying back the data from Spare RAID to the failed RAID in the background.

The advantages of the solution are:

- Eliminate the IO impact to the RAID rebuild performance;


Page 02 of 4

- Keep IO performance of rebuilding RAID be consistent;

- The write IO performance may be even better during the RAID rebuild process (if the Spare RAID has faster speed than the failed RAID).

- Though this solution brings big benefits for SSD RAID, it can be also applied to HDD RAID.



.. Description Description


Describe how your invention works Describe how your invention works,




The description to the main components:

HOST: Server or Application which needs to access the data on the SSD

Rebuild RAID: The RAID which is in the progress of rebuild or still has data i...