[32822] in North American Network Operators' Group
Re: Packet Loss
daemon@ATHENA.MIT.EDU (Tony Rall)
Thu Dec 14 12:59:03 2000
To: nanog@merit.edu
Message-ID: <OF7F047445.D02E1C30-ON882569B5.0060B0FA@LocalDomain>
From: "Tony Rall" <trall@almaden.ibm.com>
Date: Thu, 14 Dec 2000 09:50:28 -0800
MIME-Version: 1.0
Content-type: text/plain; charset=us-ascii
Errors-To: owner-nanog-outgoing@merit.edu
> I'll speculate that it occurs when packets destined for a destination do
> not get there. Most people see it via dropped ping packets..
What ping really tells you is that either the packets from A to B are being
lost or those from B to A are being lost. From the information that ping
presents there is no way to know which direction (and the routing may well
be asymmetric) is choking, nor if the problem is at one of the end points.
While I really like to use pings (with at least 100 samples) for testing
round trip packet loss, there are situations where it will give incorrect
results. If the packet loss is data, packet size, or protocol dependent,
ping probably won't tell you about the performance to be expected for the
application that you're really interested in. On top of that, some systems
will filter or throttle ICMP - this really distorts the results.
Tony Rall
