EE Department Intranet - intranet.ee.ic.ac.uk
Close window CTRL+W

ELEC97029 Distributed Computation and Networks: a performance perspective


Lecturer(s):

Aims

To teach students about network and distributed computation devices and systems (NDCDS) from the perspective of curent business, industry, service and societal activities ranging from health care to manufacturing, from commerce to security and defence.

Learning Outcomes

As computing and sensing devices and their supporting network systems become even more embedded in our societal infrastructure, it is essential to understand how they perform
at the overall system level in order to use them in the best possible manner, and to deal with the intrinsic potential and limitations of such systems. This course will provide the opportunity for students to develop an overall conceptual understanding of such systems based on probability
models and on experimentation. This understanding will allow students to determine when and how overall system level computations provide the desired outcome, and how the computation and communication components of the system interact so as to enhance or limit overall system capabilities. Aspects such as synchronisation, latency, congestion, learning and system adaptation, and security and reliability, will be examined conceptually, and enhanced by the opportunity to carry out laboratory experiments in a modern, highly distributed, large scale wired and wireless computer-communication system.

Syllabus

- Description of distributed system architectures and their components: digital sensors and actuators, processing units, local area networks, packet networks and the IP protocol, wireless ad-hoc networks. The role of protocols.
- The concept of Quality of Service (QoS). Performance metrics related to system load, response time and timeliness of data, data loss, system availability and reliability.
- Overall system modelling in terms of service requests and service units. Relation to a practical system architecture.
- System analysis in terms via experiments, probability models using analytical techniques, and simulation. Performance identities and their deterministic counterparts.
- Solution techniques for very large models. Separable solutions and product form networks.
- Solution techniques for systems with dynamic controls. G-networks.
- System adaptation to changing workloads and operating conditions. The practical use of learning and its derivation from analytical models. Gradient techniques, reinforcement learning and learning by imitation.
- Analysis of a large experimental system via theoretical models and experiments.
Assessment
Exam Duration: N/A
Exam contribution: 0%
Coursework contribution: 100%

Term: Spring

Closed or Open Book (end of year exam): N/A

Coursework Requirement:
         To be announced

Oral Exam Required (as final assessment): N/A

Prerequisite module(s): None required

Course Homepage: unavailable

Book List: