Stochastic Networks (M24)
F.P. Kelly and P.B. Key
The problem of how to control communication networks is both
mathematically challenging and very important practically, with
network operators, software vendors, network regulators and
application developers actively seeking ways of delivering new
services reliably and effectively. For example current controls in
the Internet do not scale well, and are poorly suited to deal with
new and emergent services and traffic patterns. The interplay
between the user and the network has attracted the attention of
economists as well as mathematicians and engineers.
Topics will be selected from amongst the following:
- Models of queueing networks and of loss networks (Markov
process models of networks, equilibrium distributions, closed and
open queueing networks, reversibility, loss networks).
- One-dimensional networks (Burke's theorem, Poisson flows,
independence results for queueing networks, quasi-stationary
distributions for loss networks).
- Approximations and asymptotics for loss networks (Erlang fixed
point, uniqueness under fixed routing, possible non-uniqueness
underdynamic routing, limit results for large capacities and for star
networks).
- Capacity allocation and routing strategies in queueing and
loss networks (Kleinrock's result, Gallager's algorithm, shadow prices,
decentralization, convergence considerations, analogies with
deterministic network flow).
Dynamic routing in queueing and loss networks (sticky random
routing in loss networks, Hajek's result for random graphs, state
space collapse in queueing networks).
- Random access schemes (Aloha, Ethernet, critical rates for
collision detect schemes, Aldous' result for binary exponential
back-off).
- Models of broadband networks (statistical sharing, queueing
and multiplexing models, effective bandwidths, connection acceptance,
pricing).
- Stochastic network controls for loss and queueing networks.
Loss networks: dynamic programming, reversible controls, trunk
reservation. Queueing networks: control of a single queue;
individual and social optimisation.
- Broadband controls: rate-based controls (leaky buckets)
analysis and problems; Cruz's network calculus; heavy traffic
analysis of queues in series. Multiple timescales, critical
timescales and long-range dependence.
- `Internet Networks': control of elastic traffic: window-based
flow control, modelling TCP (simple model and Ott's stochastic model
of a single link); social, network and user optimisation; prices and
tariffs as control structures; examples of control structures for
the Internet.
References
- 1.
- F.P. Kelly, S. Zachary and I. Ziedins (eds), Stochastic Networks:
Theory and Applications, RSS Lecture Note Series 4,
Clarendon Press, Oxford (1996).
- 2.
- K.W. Ross, Multiservice Loss Models for Broadband
Telecommunication Networks, Springer-Verlag, London (1995).
- 3.
- F.P. Kelly, Reversibility and Stochastic Networks,
Wiley, Chichester (1979).
- 4.
- P. Whittle, Optimization Over Time, Wiley, Chichester
(1983).
- 5.
- J.W. Roberts, U. Mocci and J. Virtamo (eds), Broadband Network
Teletraffic, Springer, Berlin (1996).
- 6.
- S.M. Ross, Introduction to Stochastic Dynamic Programming,
Academic Press (1983).