You are running multiple threads / processes (-t 200) so things are happening in parallel. The latency is how long it took for the entire request to complete from the clients point of view. So having a very busy client with lots of threads will also have an affect on this number. 

Try adjusting the thread count down  and see what happens to the latency. 

Aaron


On 06 Aug, 2010,at 07:32 AM, SSam <to_sam007@yahoo.com> wrote:


Finally I able to configure and run this program on my 3 node cluster. 
#python stress.py -n 200000 -t 200  -d 172.16.7.76,172.16.7.77,172.16.7.78  -o read
total,interval_op_rate,avg_latency,elapsed_time
83664,8366,0.0278478825376,10
145478,6181,0.0295496694395,20
177409,3193,0.027055770029,30
200000,2259,0.0238709857141,40
#

I understand that this tool prints summary results for every 10 seconds.
total,interval_op_rate,avg_latency,elapsed_time
83664,8366,0.0278478825376,10

83664 reads  in first 10 seconds,  i.e  8366 TPS ( matching with interval_op_rate)
I am not able understand  exact meaning of  avg_latency field , how it calculates this average latency.

83664 operations took 10 secs,  so average value for one operation should be (10*1000)/83664 =  0.1195 ms.

I appreciate any information on avg_latency description?

Thanks,
Sam.





#python stress.py -n 200000 -t 200  -d 172.16.7.76,172.16.7.77,172.16.7.78  -o read
total,interval_op_rate,avg_latency,elapsed_time
83664,8366,0.0278478825376,10
145478,6181,0.0295496694395,20
177409,3193,0.027055770029,30
200000,2259,0.0238709857141,40





From: Peter Schuller <peter.schuller@infidyne.com>
To: user@cassandra.apache.org
Sent: Wed, August 4, 2010 6:51:53 PM
Subject: Re: stress.py

> After I add the thift libs to CLASS_PATH , it's failing with following
> error.
>
> /stress.py: line 21: have_multiproc: command not found
> ./stress.py: line 22: try:: command not found

And that sounds like stress.py is being run by a shell rather than by
Python. Not sure how that happened since stress.py has a proper
shebang, including in cassandra 0.6.3.

Are you running it with 'python'? Is this an old cassandra that maybe
didn't have a shebang in it's py_stress (I don't know, grasping at
straws)?

--
/ Peter Schuller