Re: Best way (Fastest) to run

[pytel@xxxxxxxxxx]

When doing save to disk the most important factor is how fast your disk
subsystem can process I/Os. It depends on how data is distributed (it
influences number of seeks - the slowest disk operation).
There's no easy way to control this.
The obvious thing to do is to separate library with *SAVFs to its own ASP -
this should have the most effect.
If you have many DASD IOPs, their placement on I/O buses can also be
important. Another factor is number of DASD arms per disk IOP.
As for number of jobs which run in parallel, you can experiment - run your
mix and observe WRKDSKSTS - average disk unit utilization should not exceed
40%. When disk units are driven higher, it can actually lengthen run time.

Best regards
    Alexey Pytel



"Allen, Mark" <ALLENMA1@Mattel.com> on 03/23/99 03:27:25 PM

Please respond to MIDRANGE-L@midrange.com

To:   MIDRANGE-L@midrange.com
cc:    (bcc: Alexei Pytel/Rochester/IBM)
Subject:  Best way (Fastest) to run





We do our weekly backup disk to disk via *SAVF.  We've played around a
couple of different ways to try and speed this up.  We currently save 30
libraries each of which is initiated through a submit job to a jobq with
MAXACT=*NOMAX.  This is taking about 57 minutes.  We had been doing submit
job to a jobq set at MAXACT=3 for some of the larger libraries and
submitting the smaller ones to a jobq set at *NOMAX.  In this case it ran
about 60 minutes.

Is there a way/method I could use to speed these up.

The following is a Query I ran from the job accounting journal for these
libraries, in case it would help.

JOB                    CPU   TIME       ENTRY            AUXIO
TIME
NAME                         STARTED    TIME
ACTIVE
MUFTRANS    1,519   014504    23,501            1,460        2,997,726
INFLIBPR  30,907   014457    23,409           19,758        2,951,375
CLOCFILE04       6,033   014451    23,157            6,044        2,825,217
AADAT10   44,748   014446    23,148           36,220        2,822,103
AAOBJ10   34,045   014446    23,132           18,994        2,805,284
AASRC10   39,047   014447    22,858           25,002        2,651,021
MUFLIBPR    4,595   014503    22,454            2,960        2,391,015
INPLIBPR    9,991   014458    21,823            6,947        2,005,075
TRANSAV   27,852   014507    21,712           27,177        1,924,919
INPLIBR04 27,367   014458    21,618           17,405        1,879,129
TIMEKEEPER     11,784   014506    21,425            9,423        1,758,614
MUPLIBPR  12,003   014505    21,333            8,459        1,708,577
UTILITIES   5,898   014507    20,756            5,472        1,368,490
QGPL        4,625   014508    20,522            3,272        1,214,683
COMLIBF     9,440   014453    20,435            7,500        1,182,121
KTASSRCV8   5,649   014500    20,329            5,203        1,109,198
ACPLIB           5,248   014448    20,118            5,044          990,138
TAFLIBPR    2,650   014505    15,600            2,442          655,393
 ACFLIB               3,518   014447    15,501            4,019
613,429
 TAPLIBPR             2,254   014506    15,253            2,245
467,272
 MAOBJ10                877   014501    14,831            1,031
209,768
 MADAT10                848   014500    14,810            1,034
189,313
 MASRC10                858   014502    14,811              839
189,071
 TIMECUSTOM             604   014506    14,734              584
148,194
 MIMICSLIB              515   014502    14,657              538
114,796
 LOCLIB                 433   014500    14,619              403
79,335
 AARCV10                210   014447    14,516              199
29,488
 COMLIBP                206   014455    14,516              178
21,706
 DJFLIBR01              197   014457    14,516              108
19,631
 DJPLIBR01              155   014457    14,516               90
18,576
Mark Allen
MIS Manager
Mattel-Murray
allenma1@mattel.com

att-1.unk