Talk:Traffic engineering (telecommunications)

Not good
Traffic Engineering is not always done the way it is described in this article. The article just describes one way of doing TE, and there are many more ways. For example, consider typical MPLS Traffic Engineering which uses CSPF (Constrained Shortest Path First) to perform Traffic Engineering. The network traffic information (i.e. link bandwidth etc) is advertised and a shortest path is computed (CSPF) by pruning the links that violates constraints. In that case no statistical techniques or Queuing theory is used. Hence, the article should be written in a generalized form and different techniques should be its subsection. For example
 * MPLS Traffic Engineering
 * Traffic Engineering using OSPF weights
 * Using statistical techniques and Queuing theory

Any thoughts ? --- Faisal 15:32, 18 August 2006 (UTC)


 * Actually, I think this is a well written, concise summary description of traffic engineering. The description does not delve into how traffic is engineered, merely that it is engineered.  The article could use expansion in describing the details of the techniques involved, but the description itself is very well done.


 * Also, I take minor exception to your assertion that CSPF techniques don't use statistical techniques or queuing theory. CSPF algorithms are based on metrics that do derive, directly or indirectly, from statistical measures on the link such as bandwidth occupancy, oversubscription, queue depths, etc.  CSPF algorithms are tools to achieve engineered traffic, but very much rely on the traffic models discussed in the opening paragraph.
 * PropellerHead 15:09, 15 September 2006 (UTC)


 * Nope! CSPF could be used just with simple functions of "Residual bandwidth" and some other traffic engineering attributes on the links. They have nothing to do with queuing theory and statistical techniques. I have five publications in that area, in conference like IEEE-ICC. I suggest you to look at the work in this field. For example
 * 1. Murali Kodialam, T. V. Lakshman, "Dynamic Routing of Restorable Bandwidth- Guaranteed Tunnels Using Aggregated Network Resource Usage Information", IEEE/ACM Transactions, June 2003
 * 2. LiLi, Miliand M. Buddhikot, et al. "Routing Bandwidth Guaranteed Paths with Local Restoration in Label Switching Networks ", ICNP 2002
 * 3. Aslam, et al., "Bandwidth Sharing with Primary Paths for Restoration Routing in an MPLS Network" in Global Internet 2006
 * and dozens of more papers... --- ابراهيم 15:26, 15 September 2006 (UTC)


 * First I politely said I had a minor exception with your assertion. I also said CSPF derives directly or indirectly from statistical techniques and/or queuing theory. Even if the metric is simple - like residual bandwidth - the basis for assigning bandwidth to flows, oversubscription of the interface, etc. derive from someone sitting down and analyzing the traffic (usually with statistical models .. go figure)


 * As for reading publications ... I've been in product design, product development and network design for the past 20 years (that was long after I finished my academic career) with a number of the heavyweights in telecommunications & networking (BBN, Lucent, Nortel, Cascade, Sonus) .. I have read more publications on the subject than I care to imagine, more than one of which I've found to be pure uninformed claptrap.


 * PropellerHead 21:52, 26 September 2006 (UTC)


 * Even if the metric is simple - like residual bandwidth - the basis for assigning bandwidth to flows, oversubscription of the interface, etc. derive from someone sitting down and analyzing the traffic (usually with statistical models .. go figure). Not always right sir. You continue to talk about "offline" things ONLY. Why not to also "online" traffic engineering, with CSPF and RSVP-TE. It is a possible and highly considered in the research publications. One should not ignore it. In case it is online then the residual bandwidth change is advertised after reaching a "threshold" (to avoid advertisements overflow) and future traffic is adjusted accordingly (no operator/human involved at all). I think it is useless if you say you have done that and I say I have done that. (For example if I say that during my research we collaborate with CISCO and IETF.) It does not matter as long as I can prove what I am saying with backing of dozen of papers. And I will do that. The article could be changed (with lots of references). One cannot ignore other side of work totally and consider only "offline" stuff. ابراهيم 04:19, 27 September 2006 (UTC)