On 16/10/2012 06:44, hui jiao wrote:
there were two optimizations-redirection of reads and piggybacking of writes-which first introduced in paper "Performance Analysis of Disk Arrays under Failure" 16th VLDB conference 1990. redirection of reads: the user accesses to data that has already been reconstructed are serviced by (redirected to) the replacement disk, rather than invoking on-the-fly reconstruction as they would if the data were not yet available.
As I understand it, that's done automatically. If there is a reconstruction in progress, any read from the completed half is read from the replacement disc, but any read from the still-to-do half is reconstructed (obviously).
piggybacking of writes: the user reads that cause on-the-fly reconstruction also cause the reconstructed data to be written to the replacement disk.
No, that doesn't happen, because it would require a bitmap of which areas of the discs were reconstructed rather than a progress pointer.
have these two features been implemented in md driver now? if so, where i can find them? if not, is there any plan on them?
I can't see the second one making any significant difference in practice unless you're writing to your array so much that the reconstruction time is dramatically delayed, though maybe the paper you referred to says differently. I don't think there's a plan for it, but check out Neil Brown's blog at http://neil.brown.name/blog/ - and it's possible that you get it as a side-effect of the bad block log that's being developed/tested at the moment.
Cheers, John. -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
