Re: Bit off topic : It should go faster than this...

wizofoz2k_at_yahoo.com.au (Nuno Souto) wrote in message news:<73e20c6c.0306121631.dd8d9ff_at_posting.google.com>...
> rlro99_at_hotmail.com (Ralph) wrote in message news:<e2c49cae.0306120602.2e6a5c66_at_posting.google.com>...
> >
> > The cluster size used was 4k. As far as aggregate rate of 160mb/s that
> > is true, though I have two of these cards each with a 26 disk raid 1+0
> > array. I am copying the file from one array to the next, so one card
> > and array only seems capable of writing at 80mb/s. Or about 6mb/s per
> > spindle...thats the fastest disk I could buy....doesn't seem right to
> > me on a sequential disk write of 2gb.
>
> Hmmm, a few more points. You have 26 disk array, but on each shelf
> there are only 13 active (with one spare). And it's not clear how
> they are connected to the array controllers. There may well be
> an issue there. RAID1+0 is faster only if the "1" bit is not
> spanning the same disk and controller/shelf.
> Look into how this is setup?

Each of the 5300 controllers is connected to two enclosures by four channels. So all the '0' side of the array is in once shelf connected by two channels, the '1' side of the array is connected to the same controller using the other two channels.

According to compaqs proposal document each of these channels is capable of 160mb/s giving 320mb/s to each enclosure. According to our NT systems people there is no way to configure which channel to the enclosure controls what disks in that enclosure.

Nobody at compaq seems to be able to tell me where the bottleneck is, though they have helpfully told me that to get the best out of the controllers I could consider adding more spindles...or that I could try putting both sides of the mirror in the same enclosure...or if I copied the file to and from arrays attatched to a single controller it would be faster...

What I am trying to get across to them this that it really should go faster than this as it is and that something is broken/misconfigured etc.

Thanks again

Ralph
>
> It sounds like very little (6Mb/s) per disk. I can write-only that
> fast with my single controller, single disk box. Of course, if
> I also have to read then all heck breaks loose. Something is weird.
>
> Another thing to try is t use a larger cluster size. If the files
> you're storing are large-ish.
>
> I'm afraid ther eis no easy solution other than digging in to the
> details.
>
> HTH
>

Thanks you your help with this.

> Cheers
> Nuno Souto
> wizofoz2k_at_yahoo.com.au.nospam
Received on Fri Jun 13 2003 - 03:48:37 CDT