Just because a request takes 500ms to complete it does not mean that the other requests are waiting in line to be fulfilled (unless it is single threaded). The other requests are being fulfilled by other threads that are spawned within the JVM.
For example if your page request time under load on average was 500ms and you had 50 threads available to fulfill requests. One thread would be able to fulfill 2 req/sec. All 50 threads in the VM would be able to each fulfill 2 req/sec so 50x2=100 req/sec. The math doesn't always work out exactly this way but it is close. The number of threads you have concurrently running depends on the CPU resources you have available. You can't set active threads to 50 in a single CPU machine but in a dual or quad box it is possible. When load testing you want to try different settings for the number of active threads, while monitoring CPU and other resources. I have load tested many customer apps both CF and J2EE related and I have seen numbers well over 100 pg/sec and none of them had average page response times below 250 ms (under load). Brandon -----Original Message----- From: Matt Liotta [mailto:[EMAIL PROTECTED] Sent: Thursday, March 27, 2003 1:22 PM To: [EMAIL PROTECTED] Subject: RE: CFC scalability problems (was : RE: [CFCDev] MVCF at benorama .com) Maybe my math is screwy, but if you wanted to achieve 100pps and you had a single request that took 500ms, then you would have to be able to serve an additional 99 requests in about 5ms each. That can be done with one or more threads but the total time for all 100 requests can't be more than 1000ms. Matt Liotta President & CEO Montara Software, Inc. http://www.montarasoftware.com/ 888-408-0900 x901 > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf > Of Brandon Purcell > Sent: Thursday, March 27, 2003 11:08 AM > To: '[EMAIL PROTECTED]' > Subject: RE: CFC scalability problems (was : RE: [CFCDev] MVCF at benorama > .com) > > "Remember, to achieve 10pps you need all page requests to be handled in > 100ms or less." > > I tend to disagree with this statement, it is only true if your > application is single threaded. With a multithreaded application you can > have pg response times up around 400-500ms and still easily achieve 50-100 > pps. > > Brandon > > > > > -----Original Message----- > From: Matt Liotta [mailto:[EMAIL PROTECTED] > Sent: Thursday, March 27, 2003 10:34 AM > To: [EMAIL PROTECTED] > Subject: RE: CFC scalability problems (was : RE: [CFCDev] MVCF at > benorama.com) > > > Well if you feel that 50pps or even 10pps is enough; show me an > application with persisted CFCs that achieves that. Remember, to achieve > 10pps you need all page requests to be handled in 100ms or less. > > I invite everyone to go and test persisted CFCs and see for themselves. > See what kind of pps you can achieve and decide if that is enough for > you. > > Matt Liotta > President & CEO > Montara Software, Inc. > http://www.montarasoftware.com/ > 888-408-0900 x901 > > > ---------------------------------------------------------- > You are subscribed to cfcdev. To unsubscribe, send an email > to [EMAIL PROTECTED] with the word 'unsubscribe cfcdev' > in the message of the email. > > CFCDev is run by CFCZone (www.cfczone.org) and supported > by Mindtool, Corporation (www.mindtool.com). ---------------------------------------------------------- You are subscribed to cfcdev. To unsubscribe, send an email to [EMAIL PROTECTED] with the word 'unsubscribe cfcdev' in the message of the email. CFCDev is run by CFCZone (www.cfczone.org) and supported by Mindtool, Corporation (www.mindtool.com). ---------------------------------------------------------- You are subscribed to cfcdev. To unsubscribe, send an email to [EMAIL PROTECTED] with the word 'unsubscribe cfcdev' in the message of the email. CFCDev is run by CFCZone (www.cfczone.org) and supported by Mindtool, Corporation (www.mindtool.com).
