[Noticias-IIE] Charla del Ing. Marcelo Yannuzzi

[noticias-iie en fing.edu.uy]

Tenemos el agrado de invitarlos a la charla del Ing. Marcelo Yannuzzi,
que se realizará el lunes 20 de marzo de 2006 a las 19:00 hs en el 
Laboratorio de Software del IIE.

El título de la charla es:
On the advantages of Cooperative and Social Intelligent Route Control
y su resumen se incluye más abajo.

Marcelo es docente del IIE y está desarrollando su doctorado en la 
Universidad de Cataluña, Departamento de Arquitectura de Computadoras,
donde ha obtenido recientemente una plaza de profesor.

---------------------------------------
"On the advantages of Cooperative and Social Intelligent Route Control."

In this talk we will first review some of the main limitations of the current 
interdomain routing paradigm, including issues like:

i) slow convergence and chattiness of BGP
ii) lack of multi-path routing capabilities
iii) expressiveness and safety of policies
iv) limited Traffic Engineering (TE) capabilities
v) lack of QoS support.

  After this short review of the routing paradigm and its limitations, we will 
focus on a particular group of proposals tending to cope with most of those 
limitations. These proposals typically rely on interdomain TE strategies 
operating in relatively short timescales, and are usually referred to as 
Intelligent Route Control (IRC). IRC is being increasingly used as a way to 
optimize the cost and end-to-end performance of the outbound traffic of 
multihomed stub domains. Unfortunately, all the solutions available at present 
have in common two major drawbacks. First, all solutions are standalone, so no 
routing control interactions exist between the domains sourcing and sinking the 
traffic. The consequences of this lack of interactions are on the one hand, 
rather coarse route control over the outbound traffic of a domain, and on the 
other hand, the inability to intelligently control how traffic flows into a 
domain. The second and most important drawback is that all available solutions 
behave in a fully selfish way, that is, they operate without considering the 
effects of their decisions in the performance of the network. Indeed, it can be 
shown that when several IRCs compete for network resources, sustained traffic 
oscillations and performance degradation may occur. This is particularly 
concerning given the current trend of deployment of these solutions.
   Given these two limitations, we propose to discuss in this talk the 
advantages of extending the existing IRC model from selfish standalone to a 
cooperative and social IRC model. The keys are:

- To leverage the interactions between IRC controllers belonging to a pair of 
multihomed stub domains that exchange large amounts of traffic.
- To endow each IRC with a social route control algorithm, which adaptively 
restrains its intrinsic selfishness by learning from and evolving together with 
the network dynamics.

  Through the talk, it will be shown that that when several route controllers 
compete for network resources, the conventional ones are outperformed by those 
using a cooperative and social approach and this becomes especially noticeable 
as the network utilization increases. Indeed, extensive simulation results 
reveal that it is possible to reduce the number of path shifts by more than a 
50% on average and still obtain slightly better end-to-end traffic performance 
for delay-sensitive applications such as VoIP. Last but not least, a key 
advantage is that the extensions discussed in this talk can be installed and 
used today by simply performing software upgrades to any of the existing IRC 
solutions.