|
Graduate-level course in performance analysis and design of data networks. Emphasis is on analytical and computational methods. Topics include queueing networks; optimal routing and scheduling; and distributed algorithms.
Prerequisites: CS 338 (Computer Communication Networks), and either
ECE 434 (Random processes), or Math 366 (Applied probability),
or consent of instructor.
Text: The following are useful, but not required:
- Berstekas and Gallager, Data Networks, 2nd edition.
- Kleinrock, Queueing Systems, Volume 1
Lecture notes are available at TIS on Green Street
|
|
|
I. Issues and Models
- Network layers & architectures
- Markov & state space models
- Control
II. Introduction to Network Scheduling
- Workload & System load
- Basics of efficient scheduling
- KSRS & Klimov examples
III. Network Routing
- Workload
- Fair and efficient equilibria
- Braess' paradox
- Bellman-Ford algorithm
- Distributed algorithms
IV. Heavy Traffic Approximations
- The snap-shot principle
- Workload relaxations
- State space collapse
V. Markov Models
- Generators
- Invariance equations
- Lyapunov functions
- Limit theory
VI. The Single Queue
- The G/GI/1 queue
- Performance bounds
- Related simple models
|
|
VII. Loss Models
- Large deviations
- Effective bandwidth
- Congestion notification
VIII. Stochastic Network Models
- Markov queueing networks
- Jackson networks
- Circuit switched networks
IX. Network Stability
- Lyapunov functions
- Fluid limit models
- Multistep drift criteria
X. Network Performance
- Performance metrics
- Bounds
- Simulation
XI. Multiple Access
- Finite ALOHA model
- Infinite-station model
- Probing algorithms
XII. Optimization of Markov Models
- Optimality equations
- Dynamic programming
- Congested networks
XIII. Scheduling Revisited
- Safety stocks
- Stability
- Heavy traffic
|
|
|