Download Traffic Engineering With Traditional IP Routing Protocols by Bernard

Traffic Engineering With Traditional IP Routing Protocols by Bernard
Fortz, Jennifer Rexford and Mikkle Thorup.
Summary by Faisal Aslam
Paper Achievement: It claims that Traditional IP (IGP) shortest path routing protocols
are surprisingly effective for engineering the flow of traffic in large IP networks. It says
that global optimization of link weights is more effective then other schemes. (This
optimization is done outside the routers on some centralized location)
Traffic Engineering (TE) Definition: Improving user performance and making more
efficient use of network resources requires adopting the routing of traffic to the prevailing
demands. This task is known as Traffic engineering.
OSPF/IS-IS TE Extensions: Recent standards proposed TE extension to OSPF/IS-IS to
incorporate traffic load in the link-state advertisements (LSA) and path selection
decisions. However these extensions require modification to the routers to collect and
disseminate the traffic statistics and establish path based on the load metrics. Instead this
paper says that it is possible to select static link weights that are resilient to traffic
fluctuation and link failures, allowing the use of the traditional incarnation of OSPF and
IS-IS.
Objective function: The simple performance objective function here is to minimize the
maximum link load.
Comparison between three approaches: The author have compared three approaches
which are 1) initial configuration with unit weigthts 2) local change to the weight for
congested link 3) Global optimization of the link weights. The comparison shows that
using Global optimization of link weights one can get the best optimization.
Advantages of Global optimization: The global optimization routing (centralized
approach) has advantages like protocol stability (no unnecessary LSA generations) and
diverse performance constraints (the selection of routing parameter can be depend on
variety of performance constraints).
Traffic Engineering Framework: The author’s framework has three main steps. 1)
(Measure) The operator needs to measure the network topology and the offered traffic. 2)
(Model) evaluating possible setting of the link weights requires a way to predict how they
will affect the traffic. 3) After deciding on the values of the weights, an automated system
or a human operator needs to change the IGP configurations.
Author suggested AdvanceOSPF could handle 70% more demand than the InvCapOSPF
(weight inverse to the capabilities) and UnitOSPF (each link have unit weight).
Changing Traffic Demands: For example if the Day time traffic on and ISP is different
then night time then in such case author suggest to have single optimization for both of
the traffic demands.
Few Changes to the link weights: Sometimes it is necessary to avoid link weight
changes like situation where multiple link fails etc. But the author have analyzed that
changing a single link weight is often quite effective.