Network Traffic Engineering
FOR INFORMATION ON E-LEARNING PLEASE REFER TO THE LECTURER'S SHEET (https://corsidilaurea.uniroma1.it/user/12957).
Mondays from 11:30am to 12:30pm; DIET department 1° floor, office n°107
Outline of the Course
Traffic engineering.(4 hours) Role of and approaches to performance evaluation and network traffic engineering. Service systems: definitions and structure. Arrival and service processes. Traffic process. Performance metrics. Lindley’s recursion. Little’s law.
Scheduling and Load Balancing. (9 hours) Classification and conservation law. Priority: HOL, SJF. Scheduling: Processor Sharing and Generalized Processr Sharing, Round Robin, Credit-based scheduling, LAS. Load Balancing: push and pull policies, delay optimality, JSQ, JBT.
Packet networks. (6 hours) Jackson open queueing networks. Analysis of transit time at equilibrium. Applications to an IP network. Optimization of link capacities for a given routing. Braess’ paradox.
Cellular networks. (6 hours) Erlang’s model. Applications to cellular networks. Dimensioning of the cellular coverage under quality of service constraints. Comparison of Fixed versus Dynamic Channel Allocation.
Random access protocols. (6 hours) Slotted ALOHA: protocol, model, analysis, stabilization. Wi-Fi CSMA/CA: back-off model, saturation throughput analysis. Performance anomaly..
Congestion control. (9 hours) Fluid approximation. Examples. TCP refresher. Fluid analysis of the congestion control of a long-lived Reno TCP connection with a single bottleneck. DCTCP. Fluid analysis of DCTCP. Fairness and Network Utility Maximization (NUM). Interpretation of TCP congestion control as an adaptive, distributed controller solving NUM.
Laboratory of simulation. (20 hours). Service systems discrete simulation: event-driven and synchronous approaches. Generation of random variables. Basics of point and interval statistics. Development of simple models of stochastic simulation.
The teaching material can be downloaded from here.
- Main reference: