Not that I want to bring this subject back up, but I think there were
a few unanswered questions.

You can find the device queueing on Performance Toolkit FCX108 DEVICE
report under the "Req. Qued" field. Also is reflected in difference
between "Serv" & "Resp" (Service Time and Response Time). Shown in
example below where MDC is not enabled.

 <-- Device Descr. -->  Mdisk Pa- <-Rate/s-> <------- Time (msec) -------> Req.
 Addr Type   Label/ID   Links ths  I/O Avoid Pend Disc Conn Serv Resp CUWt Qued
 >> All DASD <<          ....       .5    .0   .2   .1  2.0  2.3  2.3   .0  .00
 80C9 3390-3 US7E42       158   4  358    .0   .2   .0  2.0  2.2  6.0   .0 1.37

For those still running VMPRF, it would be on the DASD_BY_ACTIVITY Report.

In regards to paging to 3390-9s, yes, it would be a horrible idea if it were a
real 3390-9. Otherwise, it can be done depending on the rest of the
configuration. I would avoid it, if I could. But if you are approaching the
max of 256 CP owned volumes:
1) I'd like to hear from you.
2) Your access density for paging is probably low enough that it doesn't
   matter at this point.

For those wanting to read more about Linux and PAV on VM see:
http://awlinux1.alphaworks.ibm.com/developerworks/linux390/perf/tuning_rec_dasd_PAV.shtml#begin

In regards, to the skew of access density. Traditionally, minidisk cache 
addresses
that and is why it was designed the way it was versus a private cache 
structure. But
I don't agree that it is queuing on the device level that is the bulk of the I/O
performance problems. The last 5 I/O problems I looked at were:
1 - channel contention on ESCON channels
2 - not balancing I/O across ranks
3 - not using minidisk cache
4 - PPRC config related
5 - application boo boo

For those looking to blame someone for the lack of z/VM PAV support for 
minidisks and CP
volumes, feel free to blame me; but be nice, my Mom & my manager might be 
reading this.

For those that want to discuss their systems and choices, feel free to contact 
me.
I'll respond as time permits. :-)

Bill Bitner - VM Performance Evaluation - IBM Endicott - 607-429-3286

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to