Greg Freemyer wrote:
On Thu, Sep 5, 2013 at 9:18 PM, Jan Engelhardt <jengelh@inai.de> wrote:
The current gdbm code requires that the record size be a power of 2. I have yet to see a filesystem whose ideal IO size, especially if it is derived from the fs block/sector/whatever size, is not a power of 2. But there is a first time for everything.
Are you on some odd flash filesystem, perhaps?
Linda likes raid systems and that was what she specifically said.
For raid 5 & 6 arrays in particular, the optimum write size is the stripe size. The stripe size is typically not a power of 2 (but it can be).
Bingo.
For a raid 5 or 6 writing optimally means the disk controller can implement a pure write for that stripe and the writes all happen parallel. A non-optimal write means it has to implement a read - modify - write for both the updated data blocks and for the parity block. Thus it takes roughly twice as long. It is highly worth optimizing the workload if feasible.
--- Double bingo? ;-) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org