MQSeries.net :: View topic - Shared Row

MQSIGuy · Posted: Mon Oct 04, 2010 6:47 am Post subject: Shared Row - Caching

Hi,

Can anyone help confirm or deny my theory...

I am running a broker V6 on Linux. The stack size for a flow (thread) is 8MB and I am using a shared row to cache data that is 15kb in size when viewed as a text file and this equates to roughly 800 rows - the rows have 3 columns and are small.

The flow is very small with 3 nodes and minimal ESQL so 2-4kb per node... Would caching this size of file / number of rows be worth while?

Can I add that through testing there have been no noted performance decrease using this method where as previosuly no caching was used.

Of course you may ask why then use this method? Because it is a generic flow and caching different content in different locations and it happens that just one location has this large content of data and the others have much smaller and therefore achieve better performance gains for a high throughput flow.

I appreciate I could split the flows etc removing the functionality for this example and keeping only the shared row caching for where it will be used and an advantage gained but part of this question is to simply find out what is a safe size / number of rows to use in a shared row?

Anyone done something similar? Experimented with flow stack size v caching or indeed is this even relevant???

Thanks in advance!