Ron - ZFS on the fly compression and seek

New About Yours API Help
1.2 KB, Plain text
Hi,

First off thanks for your work on this podcast, I have found it be a very useful source in my transition towards running FreeBSD as my primary OS.

In Episode 277, cogoman asked a question [1] that involved rotating a set of SSD's to prevent all of them hitting a write limit at the same time, a fairly simple solution for this would be to buy 3 SSD's set 2 of them up as a mirror, and say 6 months later attach the third to create a 3 way mirror where one drive which will mean one drive will have "fresher" flash, gaining the potential benefits cogoman desires while being simple, as well as gaining the benefits of a 3 way mirror over a 2 way. You could even buy the third later and be more or less certain the third is from a different batch too.

For my actual question, when using on the fly compression on a dataset containing some large binary files (in my case a set of binary records where each entry is of fixed length), is the ability to seek to an arbitrary point (ie the nth entry) by translating the offset to a particular sectors or does the compression mean that all accesses to these files become effectively streaming reads over the entire file, to find the data at the relevant offset.

[1] http://dpaste.com/0P0MWFC#wrap

Cheers, Ron
Pasted 2 months, 3 weeks ago — Expires in 279 days
URL: http://dpaste.com/16Y2HSR