UDB Event Monitor & Snapshot Overheads 
Author Message
 UDB Event Monitor & Snapshot Overheads
Hi,
    I would be interested in hearing from anyone who may run the Event
Monitor (any level) on a regular basis.

What sort of event types do you run?

What overheads do you  incur?

Any special changes (eg mon_heap_sz) required to run above.

Also as an alternative do you run SNAPSHOTS on a rgular basis

I am trying to work out how we may be able to collect accounting type
information (Similar to what is done on DB2 on the mainframe)

Any help appreciated,
Richard



Wed, 11 Aug 2004 18:34:48 GMT
 UDB Event Monitor & Snapshot Overheads

Richard,

We run both event monitors and snapshot monitors regularly, and notice
very little overhead.

The only event monitors we do not run constantly on all our databases are
the transaction and statement event ones : and then only because of the
volume of data they produce rather than the overhead.

I believe the difference between the UDB DB2 monitors and z/OS IFCIDs is
that UDB actually produces the information anyway, with the only overhead
being in processing it, whereas z/OS only produces the information if you
turn the IFCID on.

We also run snapshots constantly : these are done via our automated
monitoring suite of tools (from DGI : http://www.breakthroughdb2.com).

I spent quite a bit of time trying to write my own monitoring tools,
before coming across these.  They are excellent.  The money we have saved
in hardware upgrades through identifying bad SQL and fixing it has more
than paid for them in a very short space of time (less than 6 months).

I especially like SQL-GUY, which is the tool which gives you the "bad"
SQL.  The other 4 tools we have are more proactive monitoring, telling
you how the system is coping on an ongoing basis and identifying problems
which may occur and allowing you to fix them ahead of time.

If you want more info on what we do then drop me a line at

Thanks

Phil Nelson

Quote:

> Hi,
>     I would be interested in hearing from anyone who may run the Event
> Monitor (any level) on a regular basis.

> What sort of event types do you run?

> What overheads do you  incur?

> Any special changes (eg mon_heap_sz) required to run above.

> Also as an alternative do you run SNAPSHOTS on a rgular basis

> I am trying to work out how we may be able to collect accounting type
> information (Similar to what is done on DB2 on the mainframe)

> Any help appreciated,
> Richard



Wed, 11 Aug 2004 23:00:29 GMT
 UDB Event Monitor & Snapshot Overheads
Phillip,
    Thanks for the quick reply.
I am aware of the DGI tools and would very much like to get them, but we are
(not by me) committed to use IBM's TIVOLI which requires you to set up
everything yourself.

I Have been playing around  with the monitor and reckon we would run event
monitor types
DATABASE
CONNECTION
DEADLOCKS
TRANSACTION

Although may need to rethink on TRANSACTION from your comments.

I am also considering the SNAPSHOT as well.

I will then have to write some PERL script to externalise the data. Lots of
fun!

An dream of the DGI tools !

Thanks Again,
Richard

This seems to give you a bit of info to work with.

Quote:

> Richard,

> We run both event monitors and snapshot monitors regularly, and notice
> very little overhead.

> The only event monitors we do not run constantly on all our databases are
> the transaction and statement event ones : and then only because of the
> volume of data they produce rather than the overhead.

> I believe the difference between the UDB DB2 monitors and z/OS IFCIDs is
> that UDB actually produces the information anyway, with the only overhead
> being in processing it, whereas z/OS only produces the information if you
> turn the IFCID on.

> We also run snapshots constantly : these are done via our automated
> monitoring suite of tools (from DGI : http://www.breakthroughdb2.com).

> I spent quite a bit of time trying to write my own monitoring tools,
> before coming across these.  They are excellent.  The money we have saved
> in hardware upgrades through identifying bad SQL and fixing it has more
> than paid for them in a very short space of time (less than 6 months).

> I especially like SQL-GUY, which is the tool which gives you the "bad"
> SQL.  The other 4 tools we have are more proactive monitoring, telling
> you how the system is coping on an ongoing basis and identifying problems
> which may occur and allowing you to fix them ahead of time.

> If you want more info on what we do then drop me a line at

> Thanks

> Phil Nelson


> > Hi,
> >     I would be interested in hearing from anyone who may run the Event
> > Monitor (any level) on a regular basis.

> > What sort of event types do you run?

> > What overheads do you  incur?

> > Any special changes (eg mon_heap_sz) required to run above.

> > Also as an alternative do you run SNAPSHOTS on a rgular basis

> > I am trying to work out how we may be able to collect accounting type
> > information (Similar to what is done on DB2 on the mainframe)

> > Any help appreciated,
> > Richard



Wed, 11 Aug 2004 23:47:46 GMT
 UDB Event Monitor & Snapshot Overheads
Richard,

If it is any help I can share with you the Perl script I wrote to extract
the data from the output of db2evmon.  It produces fixed length files
suitable for loading to DB2 tables for further analysis.  It is
unfinished, since we bought the "Guys" before I could complete it, but
works for most event monitor types.

Phil

Quote:

> Phillip,
>     Thanks for the quick reply.
> I am aware of the DGI tools and would very much like to get them, but we
> are (not by me) committed to use IBM's TIVOLI which requires you to set
> up everything yourself.

> I Have been playing around  with the monitor and reckon we would run
> event monitor types
> DATABASE
> CONNECTION
> DEADLOCKS
> TRANSACTION

> Although may need to rethink on TRANSACTION from your comments.

> I am also considering the SNAPSHOT as well.

> I will then have to write some PERL script to externalise the data. Lots
> of fun!

> An dream of the DGI tools !

> Thanks Again,
> Richard

> This seems to give you a bit of info to work with.


>> Richard,

>> We run both event monitors and snapshot monitors regularly, and notice
>> very little overhead.

>> The only event monitors we do not run constantly on all our databases
>> are the transaction and statement event ones : and then only because of
>> the volume of data they produce rather than the overhead.

>> I believe the difference between the UDB DB2 monitors and z/OS IFCIDs
>> is that UDB actually produces the information anyway, with the only
>> overhead being in processing it, whereas z/OS only produces the
>> information if you turn the IFCID on.

>> We also run snapshots constantly : these are done via our automated
>> monitoring suite of tools (from DGI : http://www.breakthroughdb2.com).

>> I spent quite a bit of time trying to write my own monitoring tools,
>> before coming across these.  They are excellent.  The money we have
>> saved in hardware upgrades through identifying bad SQL and fixing it
>> has more than paid for them in a very short space of time (less than 6
>> months).

>> I especially like SQL-GUY, which is the tool which gives you the "bad"
>> SQL.  The other 4 tools we have are more proactive monitoring, telling
>> you how the system is coping on an ongoing basis and identifying
>> problems which may occur and allowing you to fix them ahead of time.

>> If you want more info on what we do then drop me a line at

>> Thanks

>> Phil Nelson


>> > Hi,
>> >     I would be interested in hearing from anyone who may run the
>> >     Event
>> > Monitor (any level) on a regular basis.

>> > What sort of event types do you run?

>> > What overheads do you  incur?

>> > Any special changes (eg mon_heap_sz) required to run above.

>> > Also as an alternative do you run SNAPSHOTS on a rgular basis

>> > I am trying to work out how we may be able to collect accounting type
>> > information (Similar to what is done on DB2 on the mainframe)

>> > Any help appreciated,
>> > Richard



Thu, 12 Aug 2004 01:19:34 GMT
 
 [ 4 post ] 

 Relevant Pages 

1. DB2 UDB 5.0 EEE on AIX 4.2.1 Snapshot Monitor Problem

2. Snapshot log overhead

3. How to Create a Snapshot Database on DB2 UDB on Windows

4. Correlate UDB SNAPSHOT stats WITH DB2/OS390 Accounting

5. UDB SNAPSHOT Command questions

6. UDB SNAPSHOT TABLE > REORG COLUMN

7. Event Log overflow with 18453 & 18454 events

8. UDB Monitoring with SNMP??

9. DB2 monitoring tools for UDB V7.1 AIX 4.3.3

10. DB2 monitor (list connections/get snapshot for application/dynamic sql)

11. Discussion: Snapshot Monitor API - new format

12. snapshot monitoring on 6.1


 
Powered by phpBB® Forum Software