"Model-based Identification of Dominant Congested Links"
 
In this talk, we will discuss the recently accepted paper which
proposes a model-based approach that uses periodic end-end probes to
identify whether a "dominant congested link" exists along an
end-to-end path. Here, a dominant link is considered to be a link
which causes most losses and significant queuing delays along the
path. The work describes formal definition of the dominant congested
link along with the hypothesis testing for identifying the dominant
congested links. The work also examines several model-based techniques
for identifying a dominant congested link that are based on
interpreting probe loss as unobserved delay. It presents model
identification using Hidden Markov Model (HMM) and Markov model with a
hidden dimension along with validation though simulations and
measurements.