How good is an EMC SAN?

Igor Batinic wrote:
Hi everyone...

Dirk Munk wrote:
Wilfred Pickles wrote:

Very nice commercial, now back to our regular show...

No, not a commercial. Hands on experience with all products mentioned.
I don't work for HP, and I don't sell their products. I have no reason
what so ever to make things look better than they are.

I work with IBM, EMC, HP and occasionally with some other vendors, and
those fairy tales about EVA's "user friendly" interface and some other
things are more funny than anything else.

There's really no big difference between using any of those arrays (EVA,
CX, DS4xxx). Anyone who has ever used or configured any storage array
can create arrays, LUN, storage groups (partitions) and other basic
things without looking even in the manual.

Some technical details maybe are true, but I'm not sure that all EVAs
can work in active / active mode.

Yes they can. that is true for the 'old' 3000 and 5000 versions, it is
true for the present 4000, 6000 and 8000 versions, and it will be true
for the new versions in 2008


BTW, CX can, with PowerPath, use all paths for data transfer, but only
one at the same time.

That is what I wrote, active/passive, two (or more) paths, but only the
paths to one controller can be active for one lun. This is a failover
configuration.


But your reaction reaction is typical. Managers in my company thought
I was exaggerating until I showed them it was real. A couple of weeks
ago I got a new EVA (with iSCSI option). It took me about half a day
to set it up (incl. setting up de Windows management server), without
assistence of a HP engineer. Then the project manager gave me a list
with about 4 host and 20 luns, and within 15 minutes I had them
configured.

The same thing any specialist can do on any CX or DS4xxx. That is really
not a problem.

Oh yes it is, if you want good performance.

An example from my experience with Clariion. A customer wanted to use
Clariion storage because it was cheaper than DMX storage. It was set up,
and the database performance was lousy. I was asked to have a look at
it, and I soon discovered many errors in the whole setup, which means
errors in the setup of the Clariion as well as errors in the setup of
the volume manager configuration of the SUN system.

I had the whole Clariion reconfigured, making sure that similar luns
ware created on raidgroups that were serviced by controller 0, and
raidgroups that were serviced by controller 1. Then I used the Solaris
volume manager to make stripe sets of these luns, taking care that I
used the correct interlace factor.

The end result was that the storage was 2.5 to 3 times faster then in
the first configuration.

It took about half a day work with 3 people to do this, and it took a
lot of thinking, reading, calculating and designing. With an EVA this
would have taken not even a few minutes, and no use of the volume
manager on the Solaris box.

So yes, it does make a difference.






Best regards,

Iggy