Unanswered: Significance of accurate TRANSFERRATE and OVERHEAD in Tablespace Definition
For the DB2 Engine/Optimizer to determine the best path, how important is it to the code to have accurate tablespace transferrate and overhead numbers?
The reason that I ask is that we have 3 environments: an SP2 with 11 servers (and 20 database partitions) and 2 p630 each with 5 servers (and 8 database partitions). I am also well aware that the SP2 has more internal disks and that we're underwhelmed in the p630 environments from the disk farms.
The SP2 seems to beat the p630 environments hands-down when a query is heavily IO bound.
I jsut notiiced that the tablespace definitions on the p630's have the same transferrate and overhead values as the SP2 (which probably is incorrect for the p630's).
So, aside from the fact that the SP2 has more disks, how significant is it that we probably have wrong values for the 2 paramters mentioned?
We definitely have larger memory on the p630's than the SP2 servers.
when indexs and tables are placed on different tablespaces(physically diff. drives) , optimzer uses device characersics to calcuate the cost of table scan and index scan.whichever faster will be chosen.
db2 automatically doesnt detetct these characteristics, some default values are assigned for all containers.