Domino Diva

I have found DDM probes to be very helpful in analyzing Domino performance issues with agents without having to go to all of the work to manually instrument the code.  I wrote an article for IBM System i magazine detailing how I have used DDM probes to help identify and optimize agents that are large memory consumers.  Here is a link to the article "The Importance of DDM Probes".

I hope you find the article helpful.

I was recently doing some performance work for a customer and noticed in analyzing their memory checks and memory dumps there was a memory leak.  The application that was causing the memory leak was using LS2J.  I contacted support, and they were able to confirm that the memory leak indeed existed when using LotusScript to call Java.  I checked with another customer that restarts a couple of their application servers every evening.  Sure enough, on those servers, they were executing LS2J code. 

Both of my customers are running on iSeries (AS/400, System i), and it appears this may be the only platform affected, although this is not 100% for sure.  If you use LS2J calls in your applications, I recommend you do memory dumps and memory checks to see if there is indeed a memory leak.  If you find a memory leak, here is the SPR number you can reference when calling the issue into support:

SPR #: AHOE7TPQKC  [SPR number has been corrected]
My customers that encountered the LS2J memory leak are running 7.0.3, 8.0.2, and 8.5.

In working with a customer to do some performance tuning of their Domino environment, it became very apparent that Domino release 8 requires less memory.  This particular customer is running 7 Domino servers on one LPAR on a 550 (iSeries).  Three of the Domino servers provide mail and calendaring services for about 1500 users, one server is a Sametime server providing instant messaging services, and the other three servers are used for development and testing of applications that integrate with back-end data. 

The customer was running Domino 6.5.5 prior to the upgrade.  They upgraded to release 8.0.1 on April 19.  The graph below shows a significant reduction in paging and faulting rates as soon as the upgrade happened.


The paging and faulting rates went from around 700-800 faults/second down to 300-400 faults/second.  There were no changes to the NSF buffer pool sizes for any of the servers, the only thing that changed was the release of Domino.  The graph shows pool 2, which is the base memory pool for i5/OS.  This is where all Domino servers run by default, so this was a pretty standard setup.

This significant reduction in paging and faulting would result for a customer with a similar configuration upgrading from release 7.  I have found there is very little different between release 6 and 7 servers from a memory management and consumption perspective.  Release 8 however is significantly more efficient when it comes to memory consumption.