Toad World® Forums

Running TPC-C to check on impact of increasing memory with SQL Server

I am attempting to study the impact of increasing the server memory on SQL Server. I am running BMF TPC-C with a number of users (I started with 90 & 100 virtual users), limiting the SQL Server memory to simulate server memory at 16GB, 32GB, 64GB, 96GB and 128GB.

I had expected to see better performance at the larger memory sizes, but the TPS stayed the same with different memory sizes. This tells me that the 100 users is not enough to stress the system. I got the license for 1000 virtual users and my results with 900 & 1000 are about the same (linear) with fewer users.

Is there a way to increase the load with 1000 virtual users or do I need to increase the virtual user count. Also, what is the max number of virtual users that I can use?

The data base scale is 7,000 warehouses. This is running on a Dell R910, with two CPUs (Xeon E7-4850 @ 2.0 GHz)

With the TPC-C there is a 10 virtual users per warehouse relationship that should be maintained per specification. So to stay within spec. you would need to increase you 1000 virtual user limit by increasing your licensed virtual user count. You could try to lower the latencies (delays in submitting the transactions to the server) of the transactions in the TPC-C Transaction Mix test, but lowering these values will put you outside spec. as well as may cause errors during the test run.

You could also try the TPC-E benchmark which is a replacement for the older TPC-C benchmark. It does not have any requirement of the virtual user to scale scale factor.

Thanks, I’ll try TPC-E.