Re: Three drive raid set. -- MIDRANGE-L

To me it does not look like a RAID problem per se.
53  4327    70564  74.9  131.3      5.9    47.3   83.9
54  4327    35282  74.9   50.7      6.6    13.8   36.8
55  4327    35282  74.9   70.0      5.6    22.0   47.9

I think, in this configuration the problem is because there is a 
significant imbalance in unit size.
Unit 53 is twice the size of other two.  Having set aside potential hot 
spots, normally I/O load is spread proportionally to amount of data stored 
on the drive, which again is roughly proportional to disk size.
As a result, unit 53 has to handle twice the load. If I/O load is light, 
this is not an issue, but with heavy load this unit will be simply 
overloaded.
RAID configurations are always somewhat unbalanced in size, but usually 
not so much (25% on 4-unit set, 12.5% on 8-unit set).
This configuration is simply not for heavy disk workload.

You will have to frequently rebalance data to keep data allocation 
"skewed" to compensate for disk size imbalance.
Adding one drive to the RAID set will make all drives equal in size, which 
is very very importnat for disk performance under heavy load.

    Alexei Pytel
always speaking for myself only

Where is the beginning of the end, which ends the beginning?



rob@xxxxxxxxx 
Sent by: midrange-l-bounces+pytel=us.ibm.com@xxxxxxxxxxxx
10/01/2004 04:10 PM
Please respond to
Midrange Systems Technical Discussion


To
midrange-l@xxxxxxxxxxxx
cc

Subject
Three drive raid set.






Anyone else out there have serious performance issues with a three drive 
raid set?  We are.  And we even ran STRASPBAL.  IBM tried iDoctor, etc. 
Nothing out of the ordinary.  So, their suggestion is a STRASPBAL *ENDALC 
and then add them back in with STRASPBAL *CAPACITY.

Our BP says he knows of another customer that had the same issue but just 
ordered another drive and it was solved.  Never opened pmr with IBM.

Trying to find a few "me too's", hopefully with pmr #'s, to encourage IBM 
to look harder.

We will buy more drives.  We were hoping to put it off awhile.  Boss does 
think that getting them in before year end is a good thing though, 
budgetwise.

System ASP . . . . . . . :     3316 G
% system ASP used  . . . :    74.9333

WRKDSKSTS
               Size    %     I/O   Request   Read  Write
Unit  Type      (M)  Used    Rqs  Size (K)    Rqs   Rqs 
...
50  4327    52923  74.9   15.3      8.6     7.0    8.2
51  4327    70564  74.9   25.3      7.9    13.7   11.6
52  4327    61744  74.9   29.4      6.4    17.7   11.6
53  4327    70564  74.9  131.3      5.9    47.3   83.9
54  4327    35282  74.9   50.7      6.6    13.8   36.8
55  4327    35282  74.9   70.0      5.6    22.0   47.9

The last three are the raid set in question.

Rob Berendt
-- 
Group Dekko Services, LLC
Dept 01.073
PO Box 2000
Dock 108
6928N 400E
Kendallville, IN 46755
http://www.dekko.com

--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing 
list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.