Led by Prof. Raj Srinivasan.
Traffic measurements will play a crucial role in bandwidth sizing in
MPLS networks. As the MPLS seems to be the protocol of choice for
engineering data traffic, there is a need for a methodology for
sizing Label Switched Paths (LSPs). LSPs book bandwidth along their
routes for their traffic. The amount of the booked bandwidth depends
on the traffic offered to an LSP, its volume and the QoS
requirements. The volume of traffic offered to an LSP has to be
derived from traffic measurements. There are two basic questions
that have to be answered to derive a measurement methodology. First
is what and how to measure the traffic. The measurements have to
provide information that is sufficient to measurement methodology.
This requirement is usually better satisfied with more detailed and
frequent measurements. However, taking measurements and storing them
consumes networks resources such as bandwidth and storage. As
network operators want to maximize the efficiency of their network,
they want to maximize the payload, which means that other
activities, such as measurements have to be done efficiently. An
optimal solution would be a compromise between the two requirements.
The second basic question is how to derive estimates of traffic
volume from the measurements. The measurements provide samples of
stochastic processes, which for the purpose of sizing LSPs, will
have to be turned into estimates.