On 3/9/06, Justin Greene <[EMAIL PROTECTED]> wrote: > > > -----Original Message----- > > From: Leon Rosenberg [mailto:[EMAIL PROTECTED] > > Sent: Thursday, March 09, 2006 10:28 AM > > To: Tomcat Users List > > Subject: Re: Performance tricks with multiple > > tomcat instances > > > > On 3/9/06, Filip Hanik - Dev Lists <[EMAIL PROTECTED]> wrote: > > > If I were to add my two cents, I would say the OS > > scheduler, the bus, > > > memory access all will play their part. ie, I think you > > will get more > > > juice out of 8 single CPU machines each running a tomcat, than one > > > machine with 8 CPU's, as your 8cpu machine will not have 8 parallel > > > buses, 8 drives, 8 everything etc. > > > > Of course, but it also means 8 HE rack space, 16 power > > cables, 8 switch ports etc. It also means 8 systems to > > backup, to install and to administrate. We are trying to find > > a configuration between n-one cpu servers and 1 n-cpu server > > which is best from the manageability point of view and has > > optimal tco. Personally I don't believe that anything with > > more than 2 cpu's will ever pay out, but we have to test it :-) > > I have never been a big fan of single CPU machines for threaded > applications. It it too easy for a single slow thread to botteneck the > whole system. Going beyond 2 CPUs however is expensive and I think that it > is best reserved for applications that > > - Cannot easily run across load balanced servers, such as a database > - Are written to take advantage threading in a way that will truly benefit > from having more than 2 CPUs. (i.e, we have an app where one thread is > reading data from the network and populating a queue, a thread pool is > parsing the queue data and another thread pool is processing the parsed data > so having more CPUs would allow us to increase the size of the thread pools > and thus make the app run faster).
Well, personally I had collywobbles before we've gone on 2 cpu machines, but it went well. I mean, as long as you are one cpu machine you can be sure that your threads are never really concurrent and atomic operations remain atomic (like ++) but in case of 2 cpus you start to work really concurrent.... *collywobbles* :-) > > If these two criteria are not met (and there are probably a few more I am > missing) then I think it is better to have more less powerful machines which > provided redundancy. > > I have to say that I am always surpised to read when people are setting > application servers (such as Tomcat) to maxthreads of 1000. If I had 1000 > processes running at the same time on any of my application servers it would > melt never finish any of them. It is more realistic to have 5 - 10 threads > processing at any given time and to allow requests to queue (though one must > test to find the sweet spot). I would say, you are right, but if you want to support http 1.1 keepalives 1.5 threads per user are blocked simply waiting for incoming requests, so how are you supposed to server 500 users with 75 threads? > > Just my $.02 > > Justin Leon --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
