[ previous ] [ next ] [ threads ]
 From:  Jesse Guardiani <jesse at wingnet dot net>
 To:  m0n0wall at lists dot m0n0 dot ch
 Subject:  RE: 2 x lan1641 support on m0n0wall
 Date:  Wed, 16 Feb 2005 23:48:48 -0500
Brian Watters wrote:

> Dell 2450 with 2 gig ram Dual CPU P3 866, no testing as we have seen no
> issues thus far with throughput ..

Is the maximum number of sustainable interrupts per second determined by
CPU speed, PCI bus speed, both, or other factors? Is there an equation to
determine the theoretic maximum number of sustainable interrupts on a
given system?

I'm curious because I currently run a squid cache on a FreeBSD 5.3 machine
doing about 1.6 million hits per day on a 400 mhz PII CPU with a 3c905
(xl driver) ethernet card. The cache sees a business day average of probably
1Mbit/s, peaking at full T1 throughput and troughing at just below 500kbit/s.
This machine sees about 600 to 700 interrupts per second when under full load.
Over 500 of those interrupts are generated by xl0.

How is this relevent to m0n0wall? Well, I assume that fxp cards with link0
enabled would have a higher maximum throughput due to link0's IRQ mitigation.
I have a number of fxp cards and xl cards on hand, so I could test this theory.
But how do I query the number of interrupts per second on a m0n0wall box? It has
no vmstat or systat.

And to restate the first question, how many interrupts are too many?

I hope this question is on topic. Thanks!

Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)