Take a look at the insert benchmark result, it shows almost linear increment with threads number. At last it was about 40000 rows/s with 1000 threads, is this normal?
[r...@localhost py_stress]# python stress.py -n 200000 -t 50 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 77318,7731,7731,0.00778545361112,10 > 145467,6814,6839,0.00697366528946,20 > 196434,5096,5073,0.00804287136849,30 > 200000,356,355,0.00196398434449,31 > [r...@localhost py_stress]# python stress.py -n 200000 -t 100 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 78971,7897,7901,0.0132863117957,10 > 136739,5776,5778,0.017481834951,20 > 185954,4921,4923,0.0185481736806,31 > 200000,1404,1397,0.0104579045938,35 > [r...@localhost py_stress]# python stress.py -n 200000 -t 200 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 141044,14104,14104,0.0234279322974,11 > 194018,5297,5297,0.0277147080659,21 > 200000,598,598,0.00693544078854,22 > [r...@localhost py_stress]# python stress.py -n 200000 -t 300 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 174316,17431,17511,0.0273715572676,11 > 199800,2548,2468,0.0185430930073,17 > [r...@localhost py_stress]# python stress.py -n 200000 -t 500 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 196886,19688,19691,0.0335112489779,11 > 200000,311,308,0.0150149392255,12 > [r...@localhost py_stress]# python stress.py -n 200000 -t 1000 -c 10 -d > 10.0.0.169,10.0.0.185 -o insert > total,interval_op_rate,interval_key_rate,avg_latency,elapsed_time > 200000,20000,20000,0.0264923011231,5 > On Fri, Apr 16, 2010 at 9:04 PM, Ken Sandney <bluefl...@gmail.com> wrote: > uh, yes, I am using single thread. Thank you for your link, it helps > > > On Fri, Apr 16, 2010 at 8:50 PM, Jonathan Ellis <jbel...@gmail.com> wrote: > >> sounds like you are only using a single thread. >> >> look at the second graph on >> http://spyced.blogspot.com/2010/01/cassandra-05.html >> >> On Fri, Apr 16, 2010 at 5:59 AM, Ken Sandney <bluefl...@gmail.com> wrote: >> > Hi, >> > I am just doing a simple insert test with a cluster of two nodes, but >> seems >> > relatively slow: about 1000 rows/second. The test box are normal PC, 2GB >> > RAM, Intel E3200 2.4GHz. Are there any general optimization strategy? >> > Thanks >> > >
