ADSM-L

[ADSM-L] SV: 4 spindles for db: best performing setup

2009-02-19 03:24:18
Subject: [ADSM-L] SV: 4 spindles for db: best performing setup
From: Daniel Sparrman <daniel.sparrman AT EXIST DOT SE>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 19 Feb 2009 09:21:42 +0100
Hi

First off, the number of spindles is not the only important factor. What kind 
of disk are you using for the database? Do you have a disk cache for read / 
write caching or is it JBOD in a machine?

Second off, to aquire maximum performance on a limited amount of spindles, I 
suggest you use the mirroring feature within TSM and use parallellism. That 
way, TSM will have the ability to read / write from multiple DB copies 
containing the same information. It will also protect you from a partial write, 
something that hardware mirroring / striping wont. Remember to use TSM's 
shadowing in thise case or the partial write protection is gonna be lost if you 
use parallellism.

To say what is the best setup for your particular environment isnt that easy 
since we're missing out factors such as size of database, amount of memory 
available for DB cache (bufferpool). How you treat the log will also be an 
impacting factor in your setup.

In the case of DB, RAID-5 isnt always a bad choice. DB writes/read are quite 
random. And the % of reads compared to the % of writes will also affect any 
performance gain you will get from using striping (RAID-0) since striping will 
give you a performance gain from writes, but in the case of a TSM DB not when 
it comes to reads.

In my opinion, 4 physical disks are abit low if it comes to holding both 
primary and secondary copies of the database and still have enough DB volumes. 
According to TSM performance redbooks, 6-8 DB volumes (in some cases I have 
seen number saying 8-12, but that obviously depends on the size of your 
database) is a good setup for gaining a good performance, but with only 4 
spindles, you're limited to only 4 (placing more than 1 copy on each spindle is 
gonna reduce performance if this is JBOD disks). Therefore, depending on the 
size of your database, I would say going with 2 main DB copies and 2 secondary 
DB copies would be a good choice.

If  the disksubsystem you're using doesnt have write/read cache, you might 
consider placing the volumes on a JFS-2 filesystem instead of using raw devices 
(which I usually use when I have a disksubsystem with a large amount of memory 
cache) to gain advantage of AIX's file system cache. You will however in this 
case have to adjust the MINPERM and MAXPERM settings (you adjust these settings 
with the vmo command)  for how much memory the system is allowed to use for 
filecaching to adjust your system to work optimal. Remeber not to kick up both 
your DB bufferpool and the filesystem cache since if these both start fighting 
for memory, you'll end up getting your bufferpool data or filesystem cached 
paged instead, which will reduce performance.

Best Regards

Daniel Sparrman
________________________________________
Från: ADSM: Dist Stor Manager [ADSM-L AT VM.MARIST DOT EDU] f&#246;r Oscar 
Kolsteren [Oscar.Kolsteren AT INGDIRECT.CO DOT UK]
Skickat: den 19 februari 2009 09:04
Till: ADSM-L AT VM.MARIST DOT EDU
Ämne: Re: 4 spindles for db: best performing setup

Hi Michael,

I have a similar environment and created two RAID 0 lun's of 2 disks
each. The mirroring is done in TSM itself.

I'm more interested in what segment size people use on the disks and if
a certain pp size is recommended for a volume group under AIX.

Good luck


Best Regards,
Oscar


-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Michael Green
Sent: 18 February 2009 20:23
To: ADSM-L AT VM.MARIST DOT EDU
Subject: [ADSM-L] 4 spindles for db: best performing setup

You are given 4 spindles to place TSM DB on them.
How would you set them up to squeeze maximum performance for DB?

1. Dedicate each spindle for a DBC (result: 3 DBCs, one spindle left
out, nevermind)
2. Create RAID 0 out of each pair of disks. Place a DBC on each RAID
(result: 2 DBCs on two separate RAID 0)
3. Create RAID 1 out of each pair of disks. Place a DBC on each RAID
(result: 2 DBCs on two separate RAID 1)
4. Create RAID 5 out of all spindles. Place one DBC on the RAID
(result: 1 DBC on RAID5)

Notes:
a. There is no RAID 0+1
c. Wasted space is not a concern
--
Warm regards,
Michael Green

************************************************************************
********************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom
they are addressed. If you have received this email in error please
notify the system manager.

This footnote also confirms that this email message has been swept
by an antivirus scanner for the presence of computer viruses.

ING Direct NV is a limited liability company incorporated in The
Netherlands. Registered in England & Wales, Branch Ref BR7357, 410
Thames Valley Park Drive, Reading, Berkshire, RG6 1RH.
************************************************************************
********************************************


---------------------------------------------------------------------------------------------------------------
ATTENTION:
The information in this electronic mail message is private and confidential, 
and only intended for the addressee. Should you receive this message by 
mistake, you are hereby notified that any disclosure, reproduction, 
distribution or use of this message is strictly prohibited. Please inform the 
sender by
reply transmission and delete the message without copying or opening it.

Messages and attachments are scanned for all viruses known. If this message 
contains password-protected or encrypted attachments, the files have NOT been 
scanned for viruses by the ING mail domain. Always scan attachments before 
opening them.

ING Direct NV is limited liability company incorporated in The Netherlands. 
Registered in England & Wales, Branch Ref BR7357, 410 Thames Valley Park Drive, 
Reading, Berkshire, RG6 1RH.
---------------------------------------------------------------------------------------------------------------