On 10/02/2012 02:41 PM, Pew, Curtis G wrote:
On Oct 2, 2012, at 2:04 PM, Eric Bielefeld <eric-ibmm...@wi.rr.com> wrote:
One thing no one else has mentioned is that one of the strengths of having many
disks is that you can be doing I/O on many more drives at the same time. After
all, you can only do 1 read or write from physical disk at a time. Of course,
with the huge amounts of cache, that changes. If your datacenter had 100 TB of
storage, and each physical disk stored 1 TB, you would only have a maximum of
100 physical I/O's happening at any one point in time. Since most data is not
distributed equally, probably only 20-40 I/O's would be happening at once.
Just something to think about!
With HiperPAV this is really only an issue for the back end of your storage
controller, and as you say with enough cache it shouldn't be a problem even
there.
I'm not sure I totally buy that sufficient cache necessarily eliminates
all concern about back store bandwidth. If the back end storage is
unable to sustain the average write load sent to the front end because
there are too few physical drives in the back end, wouldn't one
potentially have to have a really humongous cache, and then also have
potential concerns about how long it might take to physically de-stage
the cache data in the event of a system shut down?
--
Joel C. Ewing, Bentonville, AR jcew...@acm.org
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
