January 5, 2003 04:56 PM
(in response to DANIEL CYR)
First, remember that the CLASST, GBPOOLT and GBPCHKPT are defined
group bufferpool. So you can begin with the standard - it depends,
comes to the best settings. It depends on what time of usage is
data in the group bufferpool (will depend on the separation of your
in your virtual pools). So if you have a high rereference of pages,
do not want them to be constantly castout (which is the process
thresholds control - similar to the DWQT and VDWQT control of
virtual pools). The GBP Checkpoint acts somewhat similar to the
on a single subsystem, except it occurs for an individual GBP -
the structure owner. When it is reached, DB2 begins to castout
the changed class queues to DASD (performed by the castout owner).
set interval is one that achieves balance between the performance
frequent group bufferpool checkpoints and the impact to recovery.
the checkpoint interval the higher the resource consumption due to
frequent checkpoints. The higher the checkpoint interval the slower
recovery from a group bufferpool failure due the extended period of
between checkpoints. However, if you are duplexing the recovery
not be much of an issue. Also, you may want to double check
settings...the default is 8 minutes..so some where along the line
have been changed for some reason. Also, just like with control
writes on a
virtual pool, it is better to control the castout of changed with
(percentage of changed pages for a particular class quere) and
(percent of changed pages in GBP) thresholds.
The ratio is a setting on the group bufferpool that establishes the
of directory entries to the number of data entries in the GBP. If
we do not
have enough directory entries (entry for each page read on any DB2
only one page registered regardless of number of members with
then when a new page needs to be registered a directory will be
order to register a new page. This will then cause process
page to have to go to disk to reread and register the page.
Depending on the
number of time this occurs it can add up to significant overhead.
the -DISPLAY GROUPBUFFERPOOL command to determine how many times
occurs. IFCID 0255 can also be used to determine the number
a buffer refresh caused by cross-invalidation of a data page in the
help you get an idea of the amount of read/updates occurring among
members. The default is 5...which you could think of as 5 pages
every one 1 updated...not best in most cases.
If you need more information please feel free to email me
* Susan Lawson
* DB2 for OS/390 Certified Solutions Expert *
* IBM DB2 and S/390 Gold Consultant *
* YL&A, Inc
From: DB2 Data Base Discussion List
[mailto:[login to unmask email]On
Behalf Of Cyr, Daniel P - Wilkes Barre, PA
Sent: Sunday, January 05, 2003 12:26 PM
To: [login to unmask email]
Subject: GBPCHKPT setting
I was wondering if there is anything written about the setting of
coupling facility checkpoint frequency? What I do find is very
to tell you the truth of NO value. Currently we have BMC's Pool
it mentions the the GBPCHKPT should be set between 5 and 15.
coupling facility was brought up when DB2 v4.1 was here and the
setting (4 minutes) is still in effect.
One question, when the GBPCHKPT threshold is reached (4 minutes)
everything cast out? What happens when the threshold is
If anyone has any material on how and why to sent CLASST, RATIO,
please share it with the list.