On 5/11/2006 12:11 PM, Ron and Jenny Hawkins wrote:
I had a similar performance problem when I converted reserves on a busy RACF
primary a few years ago. Never got to the bottom of why CICS trx were
driving it so hard, but the hardware reserve was much faster than then GRS
ring in a 2xOS390 GRS ring.
RESERVE (as you probably know, Ron) is a combination of a hardware lock
and a global ENQ. So, if you do not do something else (such as using a
GRS RNL to remove the global ENQ part of the processing) a RESERVE
should not be able to be "faster" than a global ENQ.
If you -do- remove the global ENQ part of the processing, then yes, the
device-level locking done by RESERVE is faster than a global ENQ.
However, if you're just using the device-level locking, and not using
the global ENQ, then you can have one very busy system totally
monopolize access to the DASD volume for a while (possibly a significant
time period), starving all the other systems. This can occur because if
you have multiple requesters wanting shared access, the system will not
drop the hardware lock until either
(a) all of them have finished, and no one else is in the queue; or
(b) it is time to grant exclusive control to someone.
Thus, if you have a lot of sharable activity (e.g., reads to the RACF
database without true updates), one system could keep the device
reserved for possibly several minutes.
Thus, for best overall performance (of all the systems) I generally
recommend telling GRS to convert the RACF RESERVE to a purely global ENQ.
Walt Farrell, CISSP
z/OS Security Design, IBM
----------------------------------------------------------------------
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
