Yes, HP & EMC worked together on a solution, along with CDI and Bill
Lancaster.  Unfortunately, you have the 23Gb drives.  After awful
performance configured as RAID-S and then RAID-1 with the 23Gb drives, using
only 45Gb/channel, HP, EMC and CDI picked up the tab to swap  to the faster
18Gb spindles on almost 1Tb of raw storage.  Then with RAID-1 we finally
achieved fairly good throughput.

Tom G.

                -----Original Message-----
                From:   Carl McNamee
[mailto:[log in to unmask]]
                Sent:   Tuesday, April 11, 2000 8:52 AM
                To:     [log in to unmask]
                Subject:        Re: EMC on HP3k question

                Thanks to everyone for your input on this issue.  Most of
the solutions are
                similar to Terry's, although he is the first who said that
HP and EMC
                actually worked together to help.

                I think that our scenario is very close to Terry's and that
with LOT's of
                fore thought we can manage the i/o beast.

                Carl

                -----Original Message-----
                From: Terry Warns [mailto:[log in to unmask]]
                Sent: Monday, April 10, 2000 7:43 PM
                To: [log in to unmask]
                Subject: Re: EMC on HP3k question


                I have put together various EMC configurations for both test
and production
                machines.  In many cases we needed to put more than 15 EMC
logical drives on
                a controller.  We spent considerable time in designing our
user volume sets.
                On an Amisys machine for instance, we had System Volume,
Application Volume,
                and a Backup Volume.  We estimated the need for each.

                Then we started spreading the volumes across the
controllers.  For instance
                we may have chosen 16GB for our System Volume Set, so we put
1 4 GB logical
                drive on 4 controllers.  Then we may have had 180GB
Application set, where
                we would spread the 10 18GB drives over the 10 controllers.
Then we may
                have had 90GB Backup Volume and spread them over the
controllers.  We also
                tried to isolate LDEV 1 as much as possible to maximize its
IO throughput.
                We mixed and matched user volume sets with controllers.

                The result was as expected, disk IO throughput soared.  We
spent a full day
                with EMC on our configuration and a full day with our HP
CE's on our
                configuration.  With lots of white board.  We also spent a
half day
                reviewing our proposed configuration and  another full day
on our migration
                plan.

                --
                Terry Warns
                OakSoft Consulting


                Carl McNamee <[log in to unmask]> wrote in
message

news:F4B1826B1A21D211AEC5006008207AF404F885E5@dogbert.csillc.com...
                > This is a cross post from the EMC list server:
                >
                > A bit of back ground and a summary of our problem. We have
a 3700 frame
                that
                > is about 60% full of 23GB drives. The drives are split
into 4.6GB logical
                > partitions. We assigned drives to the controllers based on
HP's
                > recommendation of no more than 8-10 drives for performance
since we have
                > some very i/o intensive applications.
                > My problem is that all 32 controllers all "full", e.g.
have 8-10 drives
                > assigned, but the emc box is only half full of drives.
What I am
                interested
                > in is how many drives you have assigned to each
controller?
                > HP's theory on the 8-10 drives deals with an optimal
number of spindles,
                if
                > I'm not mistaken. Since a logical emc drive does not
necessarily equate to
                a
                > spindle we are thinking about stringing the drives out
15-20 per
                controller
                > so that we can fill the frame with drives. In our theory
we would just
                need
                > to ensure that we did not exceed the capacity of the f/w
scsi channel
                during
                > peak processing. I think that this can be done by
arranging the drives so
                > that we have a good mix of low use drives and high use
drives on each
                > controller.
                > Any thoughts or comments? Feel free to poke holes in this!
                > Carl McNamee
                > Systems Administrator
                > Billing Concepts
                >