Dear Friends,
I have been testing on a 2098-A01 processor. The system under test is an IDMS
CV running under zVSE 4.2.2 and zVM 5.4. The test has taken the form of running
on-line IDMS transactions and batch jobs and using a watch to record the
response time.
The data is on a 2105-800 shark connected by long-wave FICON channels. One test
which was undertaken produced counter-intuitive results as follows:
The 2105 was connected to the processor by one long-wave FICON channel. A
series of tests were run and response times recorded. At the conclusion of the
tests, a second long-wave FICON channel was physically connected by the IBM SSR
and varied on-line. Within 2 minutes of varying the second FICON channel
on-line, the tests of the online transactions and batch jobs was re-run.
Neither VM nor VSE were re-ipled before commencing the second test. In the test
with two channels, on-line response time and batch run times increased by
roughly 50%. Intuitively, I would have anticipated no change or perhaps a
slight drop in run times. VM minidisk cache was enabled for both tests.
I am unable to explain the variance between the two tests. I have had the SSRs
check the processor and shark and they can find no problems with the hardware.
Can anyone suggest a possible explanation?
Sincerely
Nigel Salway