Date: Mon, 22 Sep 2003 14:43:40 -0400 From: "akanwar@digitarchy.com" <akanwar@digitarchy.com> To: freebsd-hackers@freebsd.org Subject: HZ = 1000 slows down application Message-ID: <63340-220039122184340901@M2W048.mail2web.com>
next in thread | raw e-mail | index | archive | help
Hi all, =20 I am expertmenting with kernel device polling on a 4=2E8-RELEASE system=2E= =20 The application I am running is a traffic pumping application that sits in= an infinite while loop=2E At the time of this test it was doing 6Mbps in a= nd 5Mbps out traffic=2E CPU usage is 40% without polling enabled, typical CPU= usage is roughly 1/3 user, 1/3 system and 1/3 interrupt=2E I am using the = fxp driver=2E I customized my kernel with HZ=3D1000 and enabled polling via sysctl=2E=2E= =2ECPU usage dropped from 40% to 30%=2E Great so far=2E But now I noticed that my application is occassionally doing slower iterations=2E Average iteration time used to be 0=2E2 ms without polling enabled=2E With the device polling changes, the average time is still arou= nd the same, but once every few minutes the application sees iterations that are 3=2E3 seconds (*seconds*, not a typo) long=2E=20 This seems to happen as soon as I use the kernel with HZ=3D1000=2E Enablin= g or disabling device polling does not seem to make any difference to this behavior=2E I am trying to understand why there seem to be a few really lo= ng iterations=2E Could it happen that the application does not get any CPU fo= r that long? Seems very counter intuitive that higher HZ should cause this=2E= =20 Could anyone shed any light on what is happening ? Thanks, -ansh -------------------------------------------------------------------- mail2web - Check your email from the web at http://mail2web=2Ecom/ =2E
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?63340-220039122184340901>