SAN Baseline performance

Posted on 2009-12-24
Last Modified: 2013-11-14
I've been attempting to get get some baseline performance numbers for my san. That is proving to be more difficult that I thought. I've been given some tools by netapp to capture the actual performance which give me diskpersec and diskbytesperser. It's also been suggested that I use iometer to get a baseline of the performance.

I've used both, and the numbers between the two programs arent exactly matching up.

So I was wondering if anyone had any suggestions or could point me to some industry standard methodologoes for

1. Measuring Fibre SAN max performance to match up against the published maximum performance numbers.
    It's a dell AX100. Tools for performing this test would be helpful too.

2.  Measuring actual load on the san during production hours.
     The tool I have from NetApp I think is theirs, I'd like to verify their numbers.

I'm interested in the tools to perform those test, but really need to know the methodologies and real world examples.

Question by:kblackwel
    LVL 21

    Accepted Solution


    There's really no such thing like industry standard methodologies for storage performance. That's because everybody's IO needs are different.

    There's only one thing you can do, that is to measure performance with the same tool(s) and compare differences before and after whatever you're trying to do.

    What is really important is that you get a good insight in the kind of IO you get on your current san during production hours. Is it mainly random or more sequential ? How many IO/sec do you get ? What is the average IO block size ?

    Different kind of tools will generate different types of IO, so it is important that you choose (or configure) a tool to simulate IO that is as close as possible to your real world needs. E.g. if you currently have mainly random IO with 4k block size, a tool generating/measuring sequential IO with 128k block size will get you nowhere and even make you take the wrong conclusions/decisions.

    LVL 4

    Assisted Solution

    As for me, i do agree with robocat,but as you said the base line of dell AX100 would be like this.

    I/O per sec would be
    1)No.of usable disks X 120 IOPS for ISCSI
    2)No.of usable disks X 180 IOPS for FC drives

    Ex : An FC array with 12 disks
    1)Raid group1 with raid5 (4+1)
    2)Raid group2 with raid5 (4+1)
    3)Raid group3 with raid1 (1+1)

    Used disks are 4+4+1 and rest (1+1+1)3 will go for parity and mirror.

    So 9 disks X 180 IOPS = 1620 IOPS

    This array can max tolarate upto 1620 IOPS.
    You can claculate individual raid group level if required, Raid group1 can tolarate
     4 usable disks X 180 IOPS each disk = 720 IOPS max(if it is FC drive).

    If you want tools to test your IOPS on dell ax100 or other astorage arrays, then below is the link

    Performance will also be effected by cache that you have on the array, IO will be served by the cache initially , not by the disk.Latter it is written to the disk.
    So cache plays major role in performance.

    Hope this will help .........
    LVL 4

    Expert Comment

    Methodologies & Standards:
    1)What application you use Ex:exchange
    2)check for IO size that your apps uses; ex:Mostly 4k size for exchange
    3)MAX IO threshhold for the apps in peak times ; xxxIOPS , is this below required
    4)Is the IO random or sequential.; Mostly random
    5)What is the read/write ratio; ex 4:1
    6)then, from above matrix choose the raid level best suited for your apps; EX:mostly need raid 5 for data disks , and raid10 for logs.

    Raid5 for more read performance; Raid10 for better write performance; raid6 if more redendency required than performance.

    LVL 55

    Expert Comment

    Just to correct the above, RAID 10 for more read performance; not RAID 5. It's better for both read and write but it does cost more space.
    LVL 4

    Expert Comment

    In comparision to raid5, raid10 is better in write performance. Please correct me if wrong?
    LVL 4

    Expert Comment

    Riad 10 and Raid 5 are almost have same performance on read(10:9). I prefered raid5 as it is low cost when compared to raid10.
    Raid10 have twice write performnce than raid 5.

    The above values are good, if you consider same disk numbers.
    LVL 55

    Assisted Solution

    RAID 5 read performance is N-1 * IOPS of single disk, RAID 10 read performace is N * IOPS if it simply splits the read queue into two for load balancing. It is possible to improve on this, for example getting one disk to read the inner tracks and the other to read the outer tracks.

    As far as baselining goes application based performance testers such as SQLperf or jetstress give a more true to life test.

    Featured Post

    How your wiki can always stay up-to-date

    Quip doubles as a “living” wiki and a project management tool that evolves with your organization. As you finish projects in Quip, the work remains, easily accessible to all team members, new and old.
    - Increase transparency
    - Onboard new hires faster
    - Access from mobile/offline

    Join & Write a Comment

    Bellevue, WA, USA; Late December, 1980 "OK, folks, we nailed the IBM contract. We're going to have many meetings like this to discuss features and functions...and how to move from 86-DOS, the CP/M clone we just purchased, to our baby, MS-DOS. Thi…
    In this article we have discussed the manual scenarios to recover data from Windows 10 through some backup and recovery tools which are offered by it.
    This video Micro Tutorial explains how to clone a hard drive using a commercial software product for Windows systems called Casper from Future Systems Solutions (FSS). Cloning makes an exact, complete copy of one hard disk drive (HDD) onto another d…
    This Micro Tutorial will teach you how to reformat your flash drive. Sometimes your flash drive may have issues carrying files so this will completely restore it to manufacturing settings. Make sure to backup all files before reformatting. This w…

    732 members asked questions and received personalized solutions in the past 7 days.

    Join the community of 500,000 technology professionals and ask your questions.

    Join & Ask a Question

    Need Help in Real-Time?

    Connect with top rated Experts

    21 Experts available now in Live!

    Get 1:1 Help Now