Forum: Other

Post: why is my tpcc benchmark much lower than expected?

why is my tpcc benchmark much lower than expected?
msmummy
Feb 14, 2014
I am running the tpcc benchmark. I find the results only about 30% of https://forum.voltdb.com/showthread.php?8-VoltDB-tpc-c-like-Benchmark-Comparison-Benchmark-Description

For example, on a single machine (with 32 cores) I use 8 sites and 8 warehouses ( I also tried 32 sites, results are the same), with 1 asynchronous client. I get a total rate about 17,000 txn/sec. while in the above post it is suggested a 50,000 txn/sec. All transactions are run on single site.

I cannot figure out why, is my benchmark problematic? or this is a correct result? Because In the above post I see the version is v1.0, while I am testing v4.1.

I also test the latest version of h-store, and I got a result over 40,000 txn/sec.... It seems much comforting....
jhugg
Feb 14, 2014
That does sound low. Can you give a bit more information on the specific setup you are using? What hardware exactly. What is the networking between the client and the server? What specific things are you typing to make this benchmark go?
msmummy
Feb 14, 2014
the cpu is AMD Opteron 6272 (32 cores), mem is 32GB, network is Gb ethernet (but we run on single machine, so network is not used).

when i run the benchmark, i go into the $VOLTDB/tests/test_apps/tpcc, I modified the deployment.xml to set to 8 sites, and modified the number of warehouses in the run.sh.

after all these I execute "./run.sh" and "./run.sh client", then wait for the result....
msmummy
Feb 14, 2014
I tried to test version 1.0 as in the above post, but v1.0 in the github archive won't compile....