Thanks to everyone who has replied to my post.  My research has gone in a 
new direction the last few days.  

During the holiday weekend, when system activity on both of our machines 
was relatively low, I was able to compare the RMF channel utilization 
values by running the same large disk-to-disk copy batch job at each site, 
and found out that the channel utilization unumbers come out *very* 
different between the two sites! (35% utilization at one site, with no 
other batch or online users, compared to 8-10% at the other site, with 
this job *and* some production activity running.)   

So the question now is not what is causing the heavy channel utilization 
at one site...rather, it is why is the utilization showing up so 
differently between sites?  Here are some clues that we've been chasing.

1) The big difference between the two configurations is that z/OS is 
running as a z/VM guest at the site with the supposed heavy utilization, 
and running in a native LPAR at the other site.  

2) When we look at RMF Monitor III on the system with the supposed heavy 
utilization, we receive the message "CHANNEL report is incomplete - CPMF 
unavailable, modified, or restarted".

3) We have done a lot of looking since then and narrowed down the odd 
figures to only the FICON connections.

4) We brought up a native LPAR on the same machine with the VM guest, and 
RMF Monitor III does *not* show either high total utilization figures or 
the CPMF error message.

5) We looked at chpid activity from the HMC, and it was high.  However, 
the HMC manual has a disclaimer that for fiber channels, the channel is 
considered "busy" any time an operation is pending, even if the channel is 
waiting for the device to respond, and thus activity on the HMC may show 
as much higher than what is reported in RMF. 

We are hoping to have the z/OS guest isolated in its own native LPAR by 
the end of the month.  By most indications, this problem will then be 
resolved.  But I would prefer to know ahead of time what's going on here 
so that we're sure of this.  If anyone has any insight, I'd love to hear 
it.  

Thanks, Laura

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to