ADSM-L

Re: Migration Speed Has Plummeted

2005-08-17 10:16:38
Subject: Re: Migration Speed Has Plummeted
From: Joni Moyer <joni.moyer AT HIGHMARK DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 17 Aug 2005 08:58:59 -0400
Hi Richard,

My device class for the storage pools is DISK and I have read that it is
better to use a device type of FILE when using ATA disk.  Is this true and
if so, how can I correct this?  Our disk is EMC ATA Raid-3.  We thought to
get cheap disk as a landing spot for our backups and then do a quick sweep
to tape, but I'm really hurting here and we also have a data center move
coming up so this is crucial for that move.

I have the caching of the migrated files turned off.   I believe that this
is the only place that caching comes into play with TSM?   I thank you for
your help and any suggestions!!!

 Storage Pool Name                             ORACLE

 Storage Pool Type                             PRIMARY

 Device Class Name                             DISK

 Estimated Capacity (MB)                       1032192.0

 Pct Util                                      96.2

 Pct Migr                                      96.2

 Pct Logical                                   100.0

 High Mig Pct                                  70

 Low Mig Pct                                   60

 Migration Processes                           3

 Next Storage Pool                             TAPE_ORACLE

 Maximum Size Threshold                        -

 Access                                        READWRITE

 Description                                   Oracle Disk Storage Pool

 Overflow Location                             -

 Cache Migrated Files?                         NO

 Collocate?                                    -

 Reclamation Threshold                         -

 Maximum Scratch Volumes Allowed               -

 Delay Period for Volume Reuse                 -

 Migration in Progress?                        YES

 Amount Migrated (MB)                          -

 Elapsed Migration Time (seconds)              143029

 Reclamation in Progress?                      -

 Volume Being Migrated/Reclaimed               -

 Last Update Date/Time                         2005-08-16 20:30:16.000000

 Last Update by (administrator)                LIDZR8V

 Reclaim Storage Pool                          -

 Migration Delay                               0

 Migration Continue                            YES

 Storage Pool Data Format                      Native

 Copy Storage Pool(s)                          -

 Continue Copy on Error?                       -

 CRC Data                                      NO




********************************
Joni Moyer
Highmark
Storage Systems
Work:(717)302-6603
Fax:(717)302-5974
joni.moyer AT highmark DOT com
********************************



             "Richard Sims"
             <rbs AT bu DOT edu>
                                                                        To
             08/17/2005 08:35          "Joni Moyer"
             AM                        <joni.moyer AT highmark DOT com>
                                                                        cc

                                                                   Subject
                                       Re: Migration Speed Has Plummeted










Joni - Save yourself some grief and check ADSM QuickFacts for caching.
        (It's a storage pool option.)
As to the migration performance: You need to do fact gathering as
part of
analysis.  With older tapes and drives, you may be encountering
increasing
difficulty in writing blocks on the old tapes, which should be
reflected in
the AIX Error Log, if happening.  Disk transfer speed may be related
to how
it is set up (RAID type, etc.).  Disks which the OS vendor (IBM) doesn't
recognize may receive too-small queue limits.  Here is one of my AIX
notes:

Disk performance                        Disk controllers typically
afford some
                                         nice degree of parallelism
in order to
                                         improve performance/
throughput. Vendors
                                         tend to be parochail - or at
least play
                                         it safe... If you attach an
IBM disk to
                                         AIX, it sets attributes to a
nice number
                                         for Queue Depth; but if you
attach a
                                         non-IBM disk to AIX, it goes
max
                                         conservative, limiting Queue
Depth to 1,
                                         causing performance to suffer.
                                         Ref: AIX doc "Setting SCSI-
Adapter and
                                         Disk-Device Queue Limits"
                                         System performance can
essentially
                                         freeze if, for example, disk
formatting
                                         is occurring on the Shark,
and system
                                         paging space is also on the
Shark.
                                         In AIX5, some settings you
can make:
                                         1. Change maxperm, minperm
(away from
                                            their default setting of
75, 25).
                                            Possibly, lower maxperm
to 24,
                                            minperm to 12.
                                         2. Turn on I/O Pacing -
judiciously, as
                                            it can also hurt
performance.
                                         A simple way of testing disk
speed is
                                         to time how long it takes to
write a
                                         file of a given size, as via
command:
                                          time dd if=/dev/zero bs=64k
count=1000
                                           > /Some/DiskFile
                                         where the count value may be
increased
                                         as needed.
                                         See also: minperm, maxperm;
Queue Depth,
                                         disk attribute

Also, if you fall behind in migration, things just get worse as disk
fragmentation occurs, and the arm gets frantic exercise seeking all
over the
place for its next file block as a lot of distributed space on the
disk is
occupied by lingering data.  This is one of the reasons that caching
is a
performance hit.

    Richard Sims


On Aug 17, 2005, at 8:19 AM, Joni Moyer wrote:

>
> How do you know if caching is turned on?  How/where can you turn it
> off?  I
> am having lousy migration speed perfomance as well.  What can I look
> at/change?  I have an AIX 5.2 server at TSM 5.2.4.0 with LTO2 tape
> drives
> attached.  Thanks!
>