Link to home
Create AccountLog in
Avatar of mhamer
mhamer

asked on

MSA 1000 Slow settings?

currently testing an MSA1000 with 14x146 15k drives  with a view to replacingthe msa 500 we currently use on a sql server.

using diskbash  (microsoft provided disk stress)   basically copies a set file size to a drive

the msa1000 DC  is 12x slower.

both set to 100% write

also when i daisy chain another msa500 to the 1000 that performs better 10 x slower)

im new to fibre and hardware. so any pointer or how too' greatly apprciated i cant see what causing the slow down,

msa 100o is set to active active and both controllers pluged in to a single server (dl-380)



Avatar of mhamer
mhamer

ASKER

Derrrrr

just noticed the disks in the MSA100 are 10k  not 15 likle in the 500

would that make a 10 minute diffrence though in the file copy test?
MSA500 (if G2) is faster than MSA1000 since it has U320 rather than U160 backplane, mind you you won't swamp the bus if you're using random I/O. It's also a bit faster because you have read cache on the 642 that connects to it in the server.

I don't think you can connect an MSA500 to a MSA1000, the external SCSI shelves for it are MSA30 DB or SB. I don't think the MSA1000 supports direct-attached either meaning you should be going through a hub or a switch (of course you may have an integrated switch in the back).

I wouldn't expect it to be 12* slower though, are you using the same disks that were in the MSA500 or are they new ones? If new check in the ACU or ADU to see if background parity ic completed since this slows it down quite a bit.

Also if you remove one connection and disable MPIO is it faster?
Avatar of mhamer

ASKER

it is G2

but the msa 100 has 2 controllers.

msa100 does work directly (or at least it is now)

put a nother way

if you had 2 dl 585
2 dl 380
2 msa 500
1msa1000

which would you use in what config to make a 2 server cluster with better disk IO than a 500 on its own


is two 500 worth trying each with its own controller directly attatched?
Avatar of mhamer

ASKER

sorry it does have a switch built in on th eback  (ill shut up in furture :-)
ASKER CERTIFIED SOLUTION
Avatar of Member_2_231077
Member_2_231077

Link to home
membership
Create an account to see this answer
Signing up is free. No credit card required.
Create Account
Avatar of mhamer

ASKER

Great thank you


there is an eva 4 or 6000 thats on the back burner (finincial reasons)  so this was going to be an  interim solution as we had a spare 1000 and unfortunatly thought it was better thana 500.
I meant the new ones, were only anounced publically a couple of weeks ago. 4400 is faster and cheaper than the 4100.

www.hp.com/go/eva4400
www.hp.com/gp/msa2000
Avatar of mhamer

ASKER

would i be correct that the msa 500 can't make use of dual controlers (active active)  one is just a stand by?

and that a 500 g2 is quicker than an msa1000 with dual controlers that do work active active?


the it dept have been trying to get san for many a year :-) i think there has been enough £5k and £10k fixes over the years to make the purchase.

Avatar of mhamer

ASKER

also looking at quick specs a msa 1000 has more than double the ipos, doesnt that make it quicker?

500 = 14,000
1000 = 30,000
Avatar of mhamer

ASKER

IOPS even.

just to clarafyi  as it does take a while to build these tests

an MSA1000 with dual controlers can be out performed by an MSA500 g" with single controller.

Dual controller in active/active configuration may be *slower* than active passive. It's down to the write cache mirroring - in active passive mode controller 1 is active and all its write cache is mirrored to controller 2 so all of controller 1s cache is available to it. In active/active controller 2's write cache is mirrored to controller 1 and controller 1s write cache is mirrored to controller 2. That means the controllers have less available write cache since some is used to mirror the other controllers cache. So if you use active/active make sure you buy the additional cache modules.

>also looking at quick specs a msa 1000 has more than double the ipos, doesnt that make it quicker?

Only because the MSA1000 supports 42 disks whereas the MSA500 only takes 14. An MSA1000 with only 14 disks doesn't give 30,000 IOPS, more like 10,000.
Avatar of mhamer

ASKER

Thank you helped a great deal