Sorry, I meant to postpone this message in mutt and instead sent it.
On Tue, Apr 25, 2006 at 07:20:41AM -0700, Bill Moseley wrote: > This is a quite vague at this point, but I'm looking idea on how to > track down performance problems. > > We moved an application from a development machine to a client's > managed server an noticed it running quite a bit slower. Simple > requests[1] take about three to five times longer. And this happens > when the load average is above on on the fast machine and load > average is is < 0.1 on the client's server. > > I'd like to provide the ISP with more info than "It's running slow" so > I'm wondering what tools to use to compare these two machines. > > Some of the basic specs are: > > Development Client's Server > --------------- ---------------------- > CPU Athlon XP1800+ Xeon > Mhz 1150.591 1793.936 > cache 256KB 512KB > bogomips 2260.99 3565.15 > RAM 1GB .5GB > OS Deb Unstable Deb Sarge > Kernel 2.6.6 2.4.28 > Tasks 115 214 > fs xfs atime ext3 noatime, nodiratime > > It seem It seems to be CPU. Maybe the machine is taxed on the number of processes. What I notice is the machine runs at a low load average, but any small takes makes the CPU load jump. Anyway, just looking for a few recommendations on what you would do when trying to track down performance issues. > > >From that alone it would seem like the client's server would be > faster, although I'm sure that's not the entire story. Yet, the > development server can have a load average over 1 and still process > simple requests faster than the clients when its load average is < > 0.1. > > vmstat shows no swapping, so memory does not see to be a problem. > > > > > > > > > [1] It's a fast_cgi application that's just returning a small file > from the file system -- no database access involved in this request. > > -- > Bill Moseley > [EMAIL PROTECTED] > > > -- > To UNSUBSCRIBE, email to [EMAIL PROTECTED] > with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED] > > -- Bill Moseley [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]