Re: Disk Pool Volume Configuration
2005-08-17 17:59:20
Ben is absulutly correct here in my
oppinion, raw volumes are beter.
Even though aix has the best I/O perfomance
is not as good as raw volumes.
you need be a good AIX administrator
to tune the filesystem to be as good as raw. I do alsow agree with Ben
on to take earch LUN a volume and to put them in one storage pool. you
could put 2 volumes per LUN but it that depends on the perfomance on the
SATA disks versus your network load and your tape perfomance.
"It´s better to aim at perfection
and miss than it is to aim at imperfection and hit it"
Founder
Thomas J. Watson
|
|
Kær kveðja / Best regards,
Pétur
Eyþórsson · Tæknilegur ráðgjafi
eServer & Storage · Tivoli Advanced Storage Solutons Techical Expert
· MCSE · SNIA Professional ·
SNIA FC-SAN Practitioner · IBM C AIX S · TSM CP · TSRM
CP · TASS DP · IBM SAN P
Sími/tel: +354 569-7700 · GSM: +354 863-0560
www.nyherji.is
|
Ben Bullock <bbullock AT MICRON DOT COM>
Sent by: "ADSM: Dist Stor Manager"
<ADSM-L AT VM.MARIST DOT EDU>
17.08.2005 19:54
Please respond to
"ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> |
|
To
| ADSM-L AT VM.MARIST DOT EDU
|
cc
|
|
Subject
| Re: [ADSM-L] Disk Pool Volume
Configuration |
|
This is a little off the topic of your question,
but we don't
have our TSM server on a SATA SAN, just on SSA drives, so I'm not sure
what the best ways.
I see you are making all those luns into 1 LV, putting a
filesystem on that vol and then putting it in the storage pool. I would
not do it that way. I would make each LUN a volume and then put it in
the storagepool, and not aggregate them together. TSM seems to use a
round-robin approach to using volumes in a diskpool, as you can see a
single session go from vol to vol, so by aggregating them into 1 big
device, I think you kind of blow this load distribution method that TSM
uses internally, as each session will hit that one filesystem/volume and
all the disks at the same time. Yucky performance would be my guess.
In addition to that, from my experience, using the devices as
RAW volumes rather than filesystems gives you better throughput on a
typical install. If you are into VM tuning and calculating the
filesystem cacheing attributes, how to tune the read ahead buffers, etc,
etc, you can get filesystems to run better, but it takes more tinkering,
tweaking and monitoring.
I just add the raw AIX volume right into the storage pool and it
works great. i.e. here is a storagepool with a bunch volumes. Each
volume is 1 18GB disks mirrored at the OS level and then put into the
storage pool as a raw device.
tsm: TSMSERV1A>q vol stg=db_diskpool
Volume Name Storage
Device Estimated
Pct Volume
Pool Name Class Name
Capacity
Util Status
(MB)
------------------------ ----------- ----------
---------
----- --------
/dev/rstglv01 DB_DISKPOOL
DISK 17,344.0
61.8 On-Line
/dev/rstglv02 DB_DISKPOOL
DISK 17,280.0
41.4 On-Line
/dev/rstglv04 DB_DISKPOOL
DISK 17,280.0
36.6 On-Line
/dev/rstglv05 DB_DISKPOOL
DISK 17,280.0
65.9 On-Line
/dev/rstglv06 DB_DISKPOOL
DISK 17,280.0
62.8 On-Line
/dev/rstglv07 DB_DISKPOOL
DISK 17,280.0
61.0 On-Line
/dev/rstglv08 DB_DISKPOOL
DISK 17,280.0
64.2 On-Line
/dev/rstglv09 DB_DISKPOOL
DISK 17,280.0
49.4 On-Line
/dev/rstglv10 DB_DISKPOOL
DISK 17,280.0
54.6 On-Line
/dev/rstglv13 DB_DISKPOOL
DISK 17,280.0
44.8 On-Line
/dev/rstglv15 DB_DISKPOOL
DISK 17,280.0
46.1 On-Line
/dev/rstglv16 DB_DISKPOOL
DISK 17,280.0
71.9 On-Line
/dev/rstglv17 DB_DISKPOOL
DISK 17,280.0
77.7 On-Line
/dev/rstglv18 DB_DISKPOOL
DISK 34,688.0
24.4 On-Line
/dev/rstglv19 DB_DISKPOOL
DISK 17,280.0
44.8 On-Line
/dev/rstglv20 DB_DISKPOOL
DISK 17,280.0
61.5 On-Line
/dev/rstglv27 DB_DISKPOOL
DISK 17,344.0
50.2 On-Line
/dev/rstglv28 DB_DISKPOOL
DISK 17,344.0
48.3 On-Line
/dev/rstglv30 DB_DISKPOOL
DISK 17,344.0
58.7 On-Line
/dev/rstglv31 DB_DISKPOOL
DISK 17,344.0
59.5 On-Line
/dev/rstglv32 DB_DISKPOOL
DISK 17,344.0
40.9 On-Line
/dev/rstglv33 DB_DISKPOOL
DISK 17,344.0
50.1 On-Line
/dev/rstglv34 DB_DISKPOOL
DISK 17,344.0
43.4 On-Line
/dev/rstglv35 DB_DISKPOOL
DISK 17,344.0
57.9 On-Line
...
You likely won't want to do OS mirroring for the TSM DB, for other
reasons talked about in the manuals and in the archives for this
listserv, but it's the only way for me to protect from disk failure
using JBOD. In your case, if the LUNs are RAID-5 or something under the
covers, there is no need to mirror at the OS. Just create a raw volume
on each LUN and put it into the diskpool.
My 2-cents.
Ben
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Joni Moyer
Sent: Wednesday, August 17, 2005 1:28 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Disk Pool Volume Configuration
Hello,
I have just added fibre channel disk to our one AIX server. Here
are
the LUN's that were added to filesystem /tsmdev/stgpool1
CHRS044 /tsmdev/stgpool1
lun 175 100 GB
lun 178 100 GB
lun 181 100 GB
lun 184 100 GB
lun 187 100 GB
lun 190 100 GB
lun 193 100 GB
lun 199 100 GB
lun 204 100 GB
lun 210 100 GB
Total 1000 GB
What would be the best TSM volume configuration be? I have 5 different
random disk storage pools where the data resides for 24 hours before
being swept to tape. Would I create 10 storage pool volumes and put
the
appropriate number of volumes/storage pool? Or can I have more than
1
TSM volume per physical volume?
Thank you in advance!
I'm really trying to get the best performance and I'm hoping to
configure this disk properly the first time.
|
|
|