|
|
|
The I/O service time as measured by the system is the sum of a number of unique components. Each of these components can be addressed by different tuning techniques to minimize the overall response time on the device. As we discussed in Section 2.4.4.2, Technique Tutorial, we can use some simple relationships to approximate the values of each of the response time components. The DASD I/O Component Analysis Report allows you to apply this methodology to analyze the individual device response time characteristics of your systems. For this case study, we used the DASD I/O Component Analysis Report to analyze the response times of devices on an IBM 3084 with a SYSID of SYSA running MVS/XA. The peak morning periods, hours 9 and 10, were analyzed for the previous day. The CA MICS database unit that contained the installation's hardware data was denoted by database identifier B. The inquiry panel for the study is shown in Figure 2-99.
/----------------------- IOA - DASD I/O Component Analysis ----------------------\ | Command ===> | | | | Composing CA MICS Inquiry: IOA18 - DASD Component Case Study | | | | Time-span ===> DETAIL (DETAIL/DAYS) Cycle(s) ===> 01 - 01 | | Data base ===> B (PRIMARY) _ _ _ _ | | Allow Run-Time Override of Data Base ===> N (Y/N) and/or Cycle(s) ===> N (Y/N) | | | | Report by time interval ===> N (Y/N) Minimum I/O rate ===> 1 (.001 - 9999) | | | | Usage qualification processing: | | CA MICS element to test===> ________ (Any numeric HARDVA element) | | Operator to employ ===> __ (LE,LT,GT,GE) | | Numeric value to test ===> ________ (.0000000 - 99999999) | | | | Data Selection: Allow run-time override | | Date ===> _______ - _______ _______ - _______ ===> N (Y/N) | | Hour ===> 09 - 10 __ - __ __ - __ ===> N (Y/N) | | Zone ===> _ _ _ _ _ _ _ _ _ ===> N (Y/N) | | Sysid ===> SYSA ____ ____ ____ ____ ____ ===> N (Y/N) | | | | Specify extended options ===> N (Y/N) | | | \-------------------------------------------------------------------------------/
Figure 2-99. Case Study - DASD I/O Component Analysis Inquiry Panel
The DASD I/O Component Analysis Report produced by the panel
shown above is provided in Figure 2-100.
You should carefully examine a number of values on the
report, for example:
o The queue time should be proportional to the utilization
of the device. Utilizations in excess of 30 to 35
percent should be avoided, except for paging or swapping
device (see Section 2.3, ASM Analysis). High
utilization will result in significant internal queuing
for the device.
o Pending delay resulting from device reserves that are
issued by applications on other systems or controller
contention delays should be small, except for very
specialized volumes like a checkpoint volume.
o Large disconnect delays may indicate heavy seek activity
or substantial RPS misses. With multiple paths to a
device, the possibility for RPS miss is substantially
reduced. Under these conditions, heavy seek activity is
a likely culprit. In this particular case study, the
3380 devices were the standard models that can and did
experience substantial internal string contention,
leading to substantially more disconnect time than the
3390 devices, which were operating as DSLE devices with
four available paths.
o Connect time is almost completely dependent upon the
amount of data being transferred, and as such, is time
spent doing useful work.
o Delays due to ES Connection Director port busy occur
when I/O actions are delayed by the temporary
unavailability of an ESCON director port. Large ESCON
port busy delays may indicate contention for a
controller attached to an ESCON director or heavy use of
director-to-director links in a multiple director
connection.
By using the DASD I/O Component Analysis Report, you can
identify devices that are causing significant performance
problems.
CA MICS Performance Manager DASD I/O RESPONSE TIME COMPONENT ANALYSIS INTERVAL START: 06JUNyy:08:59 INTERVAL END : 06JUNyy:10:59 DURATION : 2:00:00 SYSID : SYSA ------------RESPONSE TIME COMPONENTS (MS)-------------- DEV DEVICE SSCH % ------------AVERAGE TIME (MS)------------ . 1 . 2 . 3 . 4 . 5 . NO. VOLSER TYPE RATE BUSY RESP QUEUE PEND DISC CONN ESCON 1234567890123456789012345678901234567890123456789012345 ---- ------ -------- ----- ----- ------ ------ ------ ------ ------ ------ ------------------------------------------------------- 140 WORK41 3390 5.9 9.8 10.1 1.7 1.7 3.7 2.8 0.1 QQPPDDDDCCC 142 PAG403 3390 4.4 12.7 18.9 0.0 0.2 7.3 11.5 0.0 DDDDDDDCCCCCCCCCCC 143 AAA902 3390 2.1 4.1 10.6 0.5 0.8 4.2 4.9 0.3 PDDDDCCCCC 145 SPOOL1 3390 6.2 11.6 13.2 4.4 0.1 6.2 2.5 0.0 QQQQDDDDDDCC 147 AAA904 3390 1.4 3.1 12.4 0.8 0.2 4.6 6.8 0.0 QDDDDDCCCCCCC 148 SPAR05 3390 2.5 3.9 7.5 0.1 1.8 3.0 2.6 0.0 PPDDDCCC 14C AAA917 3390 1.1 2.2 11.9 0.4 0.3 6.3 4.7 0.2 DDDDDDCCCCC 240 VIPL01 3380 3.1 7.6 26.8 2.0 0.4 17.2 7.3 . QQDDDDDDDDDDDDDDDDDCCCCCCC 241 AAA909 3380 1.3 2.2 17.7 0.2 0.7 13.1 3.6 . PDDDDDDDDDDDDDCCCC 242 PAG401 3380 3.8 15.1 40.4 0.0 1.6 26.3 12.5 . PPDDDDDDDDDDDDDDDDDDDDDDDDDDDCCCCCCCCCCCC 246 AAA923 3380 6.7 10.9 17.0 0.4 0.2 10.0 6.4 . DDDDDDDDDDCCCCCC 24B AAA914 3380 2.9 6.0 21.5 0.6 0.4 16.6 4.0 . QDDDDDDDDDDDDDDDDDCCCC 24C AAA916 3380 1.6 3.2 21.2 0.4 0.5 13.6 6.7 . DDDDDDDDDDDDDDCCCCCCC 250 PAG404 3380 4.1 14.8 36.2 0.0 0.3 24.3 11.6 . DDDDDDDDDDDDDDDDDDDDDDDDCCCCCCCCCCCC 252 WORK44 3380 5.1 11.7 26.1 1.6 1.6 14.3 8.6 . QQPPDDDDDDDDDDDDDDCCCCCCCCC 255 WORK45 3380 8.0 22.0 32.8 4.5 1.2 16.9 10.2 . QQQQPDDDDDDDDDDDDDDDDDCCCCCCCCCC 25A PAG405 3380 2.1 4.1 21.5 0.9 0.7 17.4 2.4 . QPDDDDDDDDDDDDDDDDDCC 25B WORK46 3380 2.4 6.1 28.4 1.5 1.8 14.8 10.3 . QPPDDDDDDDDDDDDDDDCCCCCCCCCC 25D WORK49 3380 2.1 13.2 65.0 0.0 0.8 55.6 8.6 . PDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD+ 25E WORK47 3380 3.6 7.9 22.7 0.2 0.7 12.7 9.2 . PDDDDDDDDDDDDDCCCCCCCCC 25F WORK48 3380 2.2 6.4 32.5 2.0 1.2 15.9 13.4 . QQPDDDDDDDDDDDDDDDDCCCCCCCCCCCCC
Figure 2-100. I/O Response Time Component Analysis Report for MVS/XA
| Copyright © 2012 CA. All rights reserved. | Tell Technical Publications how we can improve this information |