I read the performance report (url below); the performance of
3590's is impressive, but I am stuck with 9840's.
But this post isn't about tapes. What caught my eye was this note in
the document:
Note: One of TSM's tuning parameters, transaction bytes (TXNbyte),
is set at 2 GB for the purposes
of these tests. Tivoli typically recommends this setting
rather than the default setting of 25 MB.
I always thought that 25600 was the limit for txnbyte I looked in the 3.7 unix
client
manual on the tivoli web site and the limit is 25600. Then I looked in the same
manual for 4.1 and sure enough, the limit is 2097152 (= 2 gig).
So I have some questions for Tivoli. Did you annonce this change?
How are we supposed to find out about it? The unix client manual for 4.1
does not have a summary of changes section. Are we supposed to do a page
by page comparison of manuals between versions to find out what changes?
How much performance improvement is possible if I push my clients up to
2 Gig?
Where can I find a list of what else "Tivoli typically recommends"?
How come you haven't recommended this to me?
Come on guys, maybe coding the server and clients is rocket science,
but I doubt that producing the documentation is.
--
--------------------------
--------------------------
Bill Colwell
Bill Colwell
C. S. Draper Lab
Cambridge, Ma.
bcolwell AT draper DOT com
--------------------------
In <51E9450E2DA1D4118DB400508BF9AC34018A0120 AT KENMSG32.schp DOT com>, on 0In
<51E9450E2DA1D4118DB400508BF9AC34018A0120 AT KENMSG32.schp DOT com>, on
05/02/01
at 01:43 PM, "Dmochowski, Ray" <ray.dmochowski AT SPCORP DOT COM> said:
>Per the disclaimer in the paper, "performance results obtained
>in other operating environments may vary significantly" ....
>Variables outside the tape subsystem were intentionally held
>constant to demonstrate drive performance under a TSM workload.
>By making all other things equal and eliminating LAN-related
>variables, these results can 'reasonably' be considered as a
>fair representation of tape drive performance ......
>The conclusion was that
>(a) IBM employs accurate and unbiased performance
> measurements of its drives and those of its competitors
>(b) performance of IBM tape drives is markedly better than
> other drives in their classes
>(c) 3590E and 3580 LTO performance is important for sites
> with a tape subsystem that includes TSM.
>Of course, YOUR actual mileage CAN vary ...... significantly!
>-Ray
>
>-----Original Message-----
>From: sal Salak Juraj [mailto:sal AT KEBA.CO DOT AT]
>Sent: Wednesday, May 02, 2001 8:54 AM
>To: ADSM-L AT VM.MARIST DOT EDU
>Subject: AW: LTO Tape Drive Performance
>Thanks!
>Interesting reading, however, with limited usablility scope.
>This tests vere optimised highest throughputs / streaming
>(fair over all technologies, but for this single task only),
>while many of us will cope with lower throughputs because of sets of smaller
>files, network limitations etc. In such environments can tape technologies
>tested compare quite different.
>regards
>Juraj salak
>-----Ursprüngliche Nachricht-----
>Von: Dmochowski, Ray [mailto:ray.dmochowski AT SPCORP DOT COM]
>Gesendet am: Freitag, 27. April 2001 15:06
>An: ADSM-L AT VM.MARIST DOT EDU
>Betreff: LTO Tape Drive Performance
>To all ....
>A "new" performance paper by Progressive Strategies is posted on the Tivoli
>site -
>SEE ANALYST REPORTS:
>Tape Drive Performance Comparisons-Using Tivoli Storage Manager
>http://www.tivoli.com/products/solutions/storage/storage_related.html
>Ray Dmochowski
>Schering-Plough Research Institute
>Kenilworth, NJ 07033, U.S.A.
>(908) 740-3261
>ray.dmochowski AT spcorp DOT com
|