bandwidth reservation

[marbukh@antd.nist.gov]

Hi all,

In my opinion, Draft's [1] claim of the beneficial effect of bandwidth 
protection through reservation can be strengthen by referring to [2].

As it has been pointed out in [1], bandwidth reservation is beneficial for 
heavily and maybe moderately loaded networks.  However, for lightly loaded 
network complete bandwidth sharing is superior to bandwidth 
reservation.  The question remains what is the right strategy if load is 
uncertain, as in a case of default parameter setting or highly volatile 
load.  Obviously, load uncertainty creates risks of bandwidth over and 
under-provisioning.  The risk of over-provisioning is realized if bandwidth 
reservation is employed in expectation of high load, while the actual load 
is light.  In opposite case the risk of under-provisioning is realized.

Paper [2] has quantified these risks and proposed to optimize the default 
parameters by minimizing the worst-case scenario risk.  The main conclusion 
of [2] is that increase in the load uncertainty increases risk of bandwidth 
under-provisioning and thus strengthen case for bandwidth protection 
through reservation, especially in default setting context.

Some extensions of this approach to other situations of network 
provisioning under uncertain demands can be found in [3]-[5].

Appreciate any comments.

Thanks,
Vladimir

Ref.
[1]  http://www.ietf.org/internet-drafts/draft-ietf-tewg-diff-te-mar-00.txt
[2]  V. Marbukh, "Network management under incomplete information on the 
operational environment," IEEE Int. Symp. on Information Theory and its 
Applications (ISITA2000), Honolulu, Hawaii, 2000.
[3]  V. Marbukh, "Robust traffic engineering: game theoretic perspective," 
ACM SIGMETRICS Performance Evaluation Review, Vol. 30, Issue 3, Dec. 2002.
[4]  V. Marbukh, "A scenario based framework for robust network 
provisioning," International Teletraffic Congress (ITC), Berlin, Germany, 2003.
[5]  V. Marbukh, "Network provisioning as a game against nature," IEEE 
International Communications Conference (ICC), Anchorage, AK, 2003.