Regarding maximing the performance

[SAJAL B. <s.b...@qu...>]

Hi,

I want to run a test against a web server (nginx) with traffic coming from maximum possible clients , each with different IP address. I am running the Curl Loader on a physical machine with 2 dual core processors (i.e. 4 cores), and 8 GB of memory. The machine has 2 gigabit Ethernet cards. I was just wondering how to make the most efficient utilization of these resources. I have come up with a couple of options and would need some help in choosing the most appropriate to get the maximum performance.

 1.
Use the entire resources together and run the curl loader scripts with -t 4 and total number of clients to be somewhere around 150 K. Considering each client consumes around 40 KB of memory, the theoretical maximum will be (8*1024*1024) / 40 somewhere around 210 K. But since we do need some spare memory, hence we choose approx 150 K.
 2.
Divide the resources into different parts using virtualization and run one instance of curl loader on each Virtual Machine. For example. have two virtual machines each with one dual core CPU (dedicated/reserved) and 4 GB of memory (dedicated/reserved), and run one instance of curl loader on each with -t 2 and number of clients to be approximately 75 K on each machine.

First option I recon would only use one Ethernet card, but the second option would.



I also want to have a heavy load in terms of data rate or packets per second so what should be the appropriate values for TIME_URL_COMPLETION and TIME_AFTER_URL_SLEEP.



Awaiting for the response.



Thanks and Regards



----
Sajal Bhatia
Research Masters Student
QUT, Brisbane
AUSTRALIA