Toad World® Forums

TPS in TPCC and TPCE benchamrk for varrying SQL Buffer cache sizes



When I ran a 600 user workload in TPCC benchmark through quest, no matter what the buffer cache size I used at the SQL Server, I got almost the same Transactions per second for different buffer cache sizes.

However, if I run a same workload(Ex:12 users) in TPCE, with varying buffer cache sizes, I get more Transactions per second for a higher buffer cache size, than the smaller ones even though the load is same.

I feel it makes sense for the TPS to change as the buffer cache size changes in TPCE, but not able to understand why it doesn’t change in TPCC.

Is it because of the fixed standard in TPCC of every 1 user generating 0.05 TPS and no such standard of TPCE?

Has anyone seen similar behavior in case of TPCC?

Please help me understand.



Yes the TPS value staying consistent for the TPC-C regardless of the SQL Server cache size settings are more than likely due to the latency settings for the TPC-C transactions. The TPC-E does not have any latency requirements per specification, so by default BMF will submit the transactions as fast as possible. Now you can change/remove the latency settings for the TPC-C transactions by selecting the transactions in the left hand Scripts view, click on the Latency tab on the right hand properties view, and then click no delay. Note that removing these latency values will cause the workload to be outside the TPC-C specification and thus may cause other issues such as deadlocks.