Re: [squid-users] squid performance

From: Robert Collins <robert.collins@dont-contact.us>
Date: Sat, 9 Jun 2001 04:41:03 +1000

----- Original Message -----
From: "Adam Lang" <aalang@rutgersinsurance.com>
To: "Squid-Users" <squid-users@squid-cache.org>
Sent: Saturday, June 09, 2001 4:29 AM
Subject: Re: [squid-users] squid performance

> Hmm... maybe those numbers are only for shared networks.
>
> I didn't make up the 37% number. I actually got it from a book
talking
> about Cisco equipment and it was talking about ethernet.
>

Was it talking
10BaseT/Thinnet aka 10Base2/100MBit Half duplex or 100MBit full duplex?

the collision domain characteristics for a network wired with each of
the above topologies various significantly.
10Base2 did have a peak performance of ~ 37% of the transimission rate.
However 10BaseT, 100 half duplex and 100 full duplewx all make
significant improvements - the specification for interframe gap, timing,
collision detection are all different from the "original" ieee 802.3
10base 2 spec. (Don't aske me the details, I'd have to go look em up
:} )

An example: the collision rate on a 100Mbit full duplex connection
should be 0% at 96Mbit data transfer or receieve, or even 192Mbit
combined data rate (in and out).

The cisco book you had, may be dated, or talking about a specific media
type.

Rob
Received on Fri Jun 08 2001 - 12:41:10 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:00:35 MST