From owner-freebsd-hackers Tue Oct 24 11:28:05 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id LAA25160 for hackers-outgoing; Tue, 24 Oct 1995 11:28:05 -0700 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id LAA25152 for ; Tue, 24 Oct 1995 11:28:02 -0700 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id EAA27927; Wed, 25 Oct 1995 04:23:48 +1000 Date: Wed, 25 Oct 1995 04:23:48 +1000 From: Bruce Evans Message-Id: <199510241823.EAA27927@godzilla.zeta.org.au> To: bde@zeta.org.au, dennis@etinc.com Subject: Re: Async utilization..... Cc: hackers@freebsd.org Sender: owner-hackers@freebsd.org Precedence: bulk >>How much less efficient? For logins, even a local ethernet is only a few >>times more efficient than 115200 bps async through a 16450. This is >>mostly because the pty implementation is poor. >Your numbers, like unix utilization >and timing numbers, are garbage. Set up a controlled test where you know the >answer >and the numbers won't be close. FreeBSD isn't unix. The sum of the user, system and interrupt times is accurate to within 5usec * (number of context switches) under FreeBSD, but since the interrupt time is not available through any syscall and my tests involve a lot of interrupts, I just used the real time, which is accurate to about 1 part in 1000 here. My tests were only accurate to within about 10%. I knew that the numbers would be close because I knew how bad the pty implementation is and picked tests involving ptys to reduce the advantage of ethernet. Ethernet is less than 10 times as efficient as async serial on the test systems anyway (ethernet overhead on the DX2/66 is about 70% for one 10Mb/s channel; async serial overhead per 115200 bps channel is about 6.3%). Bruce