From owner-freebsd-hackers Sun Mar 21 11:14:43 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from skynet.ctr.columbia.edu (skynet.ctr.columbia.edu [128.59.64.70]) by hub.freebsd.org (Postfix) with SMTP id 8AD8914C3B for ; Sun, 21 Mar 1999 11:14:38 -0800 (PST) (envelope-from wpaul@skynet.ctr.columbia.edu) Received: (from wpaul@localhost) by skynet.ctr.columbia.edu (8.6.12/8.6.9) id OAA09025; Sun, 21 Mar 1999 14:20:52 -0500 From: Bill Paul Message-Id: <199903211920.OAA09025@skynet.ctr.columbia.edu> Subject: Re: Gigabit ethernet revisited To: dillon@apollo.backplane.com (Matthew Dillon) Date: Sun, 21 Mar 1999 14:20:51 -0500 (EST) Cc: julian@whistle.com, hackers@freebsd.org In-Reply-To: <199903210730.XAA09427@apollo.backplane.com> from "Matthew Dillon" at Mar 20, 99 11:30:42 pm X-Mailer: ELM [version 2.4 PL24] Content-Type: text Content-Length: 2333 Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Of all the gin joints in all the towns in all the world, Matthew Dillon had to walk into mine and say: > You need to do the test I suggested sinking the packets into the > bit bucket with ipfw. Considering that these machines are at work and I'm at home, I don't really want to take the chance of fat-fingering something like this and taking the machine off the net. I'll have to see if I can try this on Monday. Again, I already know the receiving NIC is getting the packets and that they're making it as far as udp_input(). > You also need to do a cpu loading test - graph the cpu utilization > as displayed by vmstat or 'systat -vm 1' verses the packet load, > both going into the bit bucket, or being read by a process. > > If there is any cpu available, runnable processes will run no matter > what the cpu load. Like I said earlier: the machine is very loaded with interrupts. According to systat -iostat, about 70% of the CPU is going to interrupts, 30% to system usage and almost nothing to 'user' usage. Again, this is with a UDP test. Now, in case I didn't make it clear, there is _NO_ other activity on the system: I am using sinkmode with ttcp, meaning that I am not reading files from the disk. There is no disk activity at all, only network activity (there's a tiny amount of activity on the xl0 interface too since that's where I'm logged in). On the transmit machine, most of the CPU is going to the system. With TCP, things look much different: systat -iostat shows the receiving machine 30% idle, with the rest of the CPU usage split fairly evenly between interrupt processing and the system. However there is still almost no 'user' activity. I can get 55MB/sec over TCP like this, which doesn't suck, but it should be better. -Bill -- ============================================================================= -Bill Paul (212) 854-6020 | System Manager, Master of Unix-Fu Work: wpaul@ctr.columbia.edu | Center for Telecommunications Research Home: wpaul@skynet.ctr.columbia.edu | Columbia University, New York City ============================================================================= "It is not I who am crazy; it is I who am mad!" - Ren Hoek, "Space Madness" ============================================================================= To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message