主 题: Dynamic Scheduling of Open Multi-class Queueing Networks in a Slowly
报告人: 常俊霞 博士 (Georgia Institute of Technology)
时 间: 2005-12-13(周二) 下午 10:30 - 11:30
地 点: 理科一号楼 1418
We provide a uniform approach to study the scheduling policies of high
speed computer communication networks that has a bursty traffic
pattern. We model such networks as open multiclass queueing networks in
a randomly and slowly changing environment. At each state of the
environment, the system operates as an open multiclass queueing
network. We show that this system can be approximated by its fluid
analog: a stochastic fluid model (SFM).Given the SFM solution, we
provide a translation method to derive a suitable scheduling policy for
the original system. We show that this derived scheduling policy is
asymptotically optimal if the given SFM solution is optimal. Such type
of method is referred to as a successful translation method for fluid
model solutions. Note that current literature reveals that it is
considered difficult to provide a general successful translation
method, even for the case without a ``changing\\" environment. In this
study, we provide such a general method! for both cases, with and
without a changing environment. Through this method, we establish a
framework to search for near-optimal scheduling policies for open
multiclass queueing networks.
