Zartash Afzal Uzmi
Associate Professor, CS/EE
Lahore University of Management Sciences (LUMS)
D.H.A., Lahore, 54792 Pakistan
Phone: +92 35608202
Fax: +92 42 5898315
Email: myfirstname AT lums DOT edu.pk

Research

My general interests are in Communications and Networking areas. Throughout my career, I have been involved in projects that fall in these two closely related areas. Over the past few years, my research and other development projects that I worked on have focused on various entities used in a communication network, from physical layer all the way up to application layer. More specifically, I am interested in the following areas, in no particular order. [contact me for additional details about these areas]

Reliable Networks

The Internet is inherently inferior to the plain old telephone network in terms of providing guaranteed delivery of data. Furthermore, statistical multiplexing in the Internet does not allow us to promise fixed values for delay and delay variations. The situation is exacerbated when a failure occurs somewhere on the Internet; recovery from such a failure is usually a slow process. As a result, it is hard to promise "quality of service" on the Internet. In our study of reliable networks, we see if it is possible to guarantee bandwidth and delay values on the network under steady conditions, when nothing fails. We further question if it is possible to provide "fast" recovery when network does in fact suffer from the failure of a link, a node, or a group of links and nodes.

Related web pages: 1) Traffic Engineering

Wide Area Networking Systems

Protocols and network architectures are continuously evolving for carrying network traffic over local and wide area networks. For example, Ethernet, which was primarily considered a local area network technology, is seeing large deployments in service provider networks over wide areas, owing to its low cost and ubiquity. The traditional control plane for Ethernet is based on the spanning tree algorithm (STA). While STA makes an excellent control plane for local area Ethernet deployments, service provider deployments over wider areas require alternate control plane architecture, such as GMPLS. While this debate on using GMPLS as the control plane for Ethernet in wider areas continues, there is room to model and analyze such a mechanism. Other problems of wide area networking systems include finding efficient routing policies and QoS provisioning across multiple autonomous systems.

Related web pages: 1) GELS, 2) Traffic Engineering

IP Route Aggregation

IP routers determine the next hop of each IP datagram they receive by looking up the destination IP address in the routing table using Longest Prefix Matching (LPM). The speed of looking up a destination IP address using an LPM scheme depends upon the number of prefixes in the routing table. Existing LPM schemes use the routing table as is provided by the routing protocols without any attempt to reduce the number of entries in the routing table by means of aggregation. We try to answer the following questions: 1) To what extent we can reduce the number of prefixes in the routing table by means of aggregation without sacrificing correctness in routing? 2) Can this reduction make the process of longest prefix matching faster, and to what extent? 3) Can updates received for the original routing table be efficiently incorporated in the aggregated routing table?

Multihop Cellular Relays

Wireless data in seeing an explosive growth in terms of the number of subscribers as well as the data rates. Most wireless data systems use a cellular architecture whereby a base station is placed at the "center" of a region called "cell", while mobile subscribers are scattered around the base station. Typically, a city is serviced by many such cells adjacent to each other. There is another trend of using cellular architecture to serve wireless voice and data to static users in their homes or offices by means of what is known as fixed wireless access. In such case, the cellular architecture is inherently unfair to users that happen to be far from the base station at the edge of the cell. To maintain data rates above a certain threshold throughout a cell, use of relays has been proposed. A quantitative study on the impact of relays and cost-benefit analysis is what is still needed.

Sensor and Ad Hoc Networks

Sensor and Ad Hoc networks research gained popularity in the last decade. Researchers have looked at sensor networks from various viewpoints. We notice that researchers contributing to sensor networks have vast backgrounds ranging from information theory, communication theory, signal processing and control theory to networking and computer science. I have previously looked at the naming issue and MAC protocol design in a mobile environment. I am also interested in determining asymptotic solutions to coverage and connectivity problems in a sensor network.

Signal Processing Applications

Signal processing is becoming a generic field and is finding its applications in all sorts of scenarios. In the past, we have looked at signal processing algorithms for video coding and communications (multiuser detection and code acquisition in CDMA systems).

Sponsors

My research activities have been sponsored by various agencies including:

Interested in working with me?

If you want to do research work with me, please make sure that your interests match mine before contacting me. In addition to this, I primarily work with students who are already enrolled or intend to enroll at LUMS - hey, I get paycheck from them. If you are a student somewhere else, please try to find a local advisor.