[28678] in North American Network Operators' Group
Re: Optical Crossconnects and IP
daemon@ATHENA.MIT.EDU (Tony Li)
Thu May 11 22:40:50 2000
Date: Thu, 11 May 2000 19:38:17 -0700
Message-Id: <200005120238.TAA02977@miata.procket.com>
From: Tony Li <tli@procket.com>
To: alex@genesyslab.com
Cc: avg@kotovnik.com, akyol@pluris.com, tli@procket.com,
nanog@nanog.org
In-reply-to: <064401bfbbb6$1a08f5a0$b608a8c0@alex1.genesyslab.com>
Errors-To: owner-nanog-outgoing@merit.edu
| In technology, an attempts to calculate traffic floods _before_ real
| congestion appear cause a lot of useless work - and sometimes simple
| models (like liquid etc) are working best.
Correct. Most traffic engineering work uses historical data to determine
where the hot spots already are. No prediction is necessary.
Given the law of large numbers and the periodicity of Internet traffic,
this seems to be a very good predictor of future behavior.
Of course, this is congestion in the large, with only low frequency
components. It says nothing about high frequency congestion.
Tony
