jiri kulhan

Hi,

I do not quite understand the relationship between the above. The common
practice in case of TRANS2BIG is to increase number of global buffers
(when the application logic cannot be changed). That's OK.
But the GT.M do***entation also says that greater block size could
negatively influence transaction processing (restarts etc.). At this
point I'm getting lost.

To me, greater block size basically means that smaller number of blocks
gets 'dirty' when an update is performed, so the cache utilization is
lower (in percentage, not in absolute memory size). So for greater block
size the cache should actually accomodate bigger number of updates
(given that number of global buffer blocks and global mapping remains
the same).

What is wrong with this assumption? Am I just mixing two separate things
(TRANS2BIG and restarts) together? Does it mean that block size affects
restarts, cache size transaction size, but there is no relation between?

Thanks,
Jirka