On Tue, Aug 12, 2008 at 10:54:43AM -0500, Barton Wold wrote: > I recently changed jobs and have found that my new place uses Veritas > striped volumes that go to a hardware array that is stripe/mirror'd. > There are a lot of smart people here who have done benchmark testing > to confirm that this is faster (I haven't seen those results). This > doesn't make sense to me. Wouldn't it be slower?
Not necessarily. > I would think that > you would be trying to write to different parts of the same drive at > once?!? Perhaps, but even if true, you can't say that that specific effect outweighs others without testing. On some systems, the benefit of scheduling multiple LUNs for I/O on the host (sw striping) dramatically improves IO, even though the HW striping may be more efficient. Array level caches may negate any drawbacks of writing to a LUN in different locations. Also, it depends on the specific array and layout. If there's a large number of physical disks, then the stripe/stripe may not be doing double writes to any cylinders. Only some layouts would have that particular problem. -- Darren _______________________________________________ Veritas-vx maillist - Veritas-vx@mailman.eng.auburn.edu http://mailman.eng.auburn.edu/mailman/listinfo/veritas-vx
