From owner-freebsd-hardware Mon Sep 16 10:46:25 1996 Return-Path: owner-hardware Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id KAA21020 for hardware-outgoing; Mon, 16 Sep 1996 10:46:25 -0700 (PDT) Received: from saguaro.flyingfox.com (saguaro.flyingfox.com [204.188.109.253]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id KAA20998; Mon, 16 Sep 1996 10:46:21 -0700 (PDT) Received: (from jas@localhost) by saguaro.flyingfox.com (8.6.12/8.6.10) id KAA07081; Mon, 16 Sep 1996 10:44:55 -0700 Date: Mon, 16 Sep 1996 10:44:55 -0700 From: Jim Shankland Message-Id: <199609161744.KAA07081@saguaro.flyingfox.com> To: jab@rock.anchorage.net, kelly@fsl.noaa.gov Subject: Re: Very Slow Ethernet Link Cc: freebsd-hardware@FreeBSD.org, hackers@FreeBSD.org Sender: owner-hardware@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk OK, so several correspondents have pointed out that 0.356ms (ping RTT for 127.0.0.1 on FreeBSD) is less than 1.1 ms (ping RTT for 127.0.0.1) on Linux. On the assumption that the original mail was not a mis-timed April Fool's prank, let me make two small, additional observations: (1) Packets to 127.0.0.1 will go through the loopback interface, so Ethernet has nothing to do with it. (2) It is possible that jab@rock.anchorage.net was looking at the ttl value rather than the RTT. The ttl was 64 for Linux, 255 for FreeBSD. This has nothing to do with ping times, but rather with the maximum number of hops that a packet can make before it is discarded on the assumption that there is some routing loop. It needs to be larger than the largest number of hops that a packet could legitimately make en route from its source to its destination. 64 is probably sufficient in today's Internet; 255 works, too, and may be a better choice. If ttl is to be the measure of performance, then sysctl -W net.inet.ip.ttl=32 will cause FreeBSD to run circles around Linux :-). Jim Shankland Flying Fox Computer Systems, Inc.