> From: Christopher Schultz [mailto:[EMAIL PROTECTED]
> wicket0123 wrote:
> | JMeter reports that for 500 concurrent users making request to our
> | application, the average response time was 1 second. That already
> | broke our SLA which is 15 milliseconds.
I presume you've already done such obvious checks as ensuring all the cores on
your CPUs are maxed out at that load, rather than it being due to (say)
database or network latency, or the network connection between the test boxes
being saturated? 500 concurrent connections, 1 second response time = 1/500
sec of network bandwidth per request. On 100 Mbit/s Ethernet (call it 10
Mbyte/s), that's 20 kbyte per response assuming you can get your network 100%
loaded - which is an impossible goal. What's your request and response size?
Are you *sure* your network is behaving at the speed you think it is? What
happens if you write a little TCP/IP app that just pipes bytes from server to
client?
I presume you've also run with smaller loads, to determine a profile of when
the problems start? 400 users is OK, I assume, or you wouldn't be testing with
500 - you'd be trying to find out what was wrong with 400.
> If I ping www.google.com from my home computer, I get a
> 14.593ms average roundtrip time.
Lucky you. It's rare I see under 200.
I do wonder about the OP's requirements. What's driving them?
> | 1) out of that 1 second, how much was due to network?
> If you do localhost testing, you could estimate that.
... but, if you've maxed out all your CPUs with Tomcat, be aware that your
testing software will reduce the available throughput for Tomcat.
- Peter
---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
