| This section describes the highlights of recent research done by my research group (Department of Elec. Commun. Engg. Indian Institute of Science) in the areas of communication networks and multimedia systems.
I am interested in designing, building, analyzing, and measuring networked systems that are composed of multiple autonomous, potentially untrusted entities. There are numerous examples of such systems: the Internet routing infrastructure, public wireless networks, peer-to-peer systems, large-scale distributed computing infrastructures (e.g., grid computing), Web services, content delivery networks, and enterprise networks.
These systems present three challenges to system designers:
1. Developing communication protocols that ensure correct, predictable, and robust operation;
2. Designing practically deployable techniques for secure operation; and
3. Supporting fault diagnosis, troubleshooting, and monitoring.
Some of the research works investigated are:
Agent Based Schemes to Resolve Some of the Problems in Multimedia communications
The rapid development in internet technology and the growing number of users of internet
are demanding new services that are based on multimedia applications. Multimedia
applications are bursty in nature and require real time services from communication
network. There are several issues to be resolved in multimedia communications such
as, digital representation of media, coding and compression, reliability, storage and
retrieval techniques, human computer interaction design, QoS (quality of service )
management such as bandwidth, delay and jitter, QoS routing, media synchronization
and playout, security, and a real time operating system at the hosts.
Looking at the agent technology and applications in various fields (E-commerce, information retrieval, etc.,) and the advantages such as flexibility, customization, adaptability, we planned to investigate and explore the use of static and mobile agents in resolving some of the problems in multimedia communications. The problems that have been tackled using agent technology are network monitoring to aid in network resource management, QoS management at network and application level during dynamic network conditions, QoS routing to increase acceptance ratio of applications, streams synchronization and playout and its application in Internet-based education systems. Further we are planning to work on adaptive media playout.
We observed by simulation that agent technology offers several advantages: flexibility; adaptability; maintainability; reusability; decomposition of complex tasks; autonomous and asynchronous execution; knowledge sharing; easy to incorporate intelligence; encapsulates user defined protocols; network traffic reduction; provides real world solutions. But there are some problems that are important to be resolved in the agent technology such as mobile agent platform creation, agent transfer protocols, addressing, portability, security to hosts and vice versa.
Applications of Neural Networks in Mobile Networks
With the advancement in mobile communication networks and portable computing
technologies, the transport of multimedia applications over the wireless channel is
challenging due to the severe resource constraints of the wireless link and mobility.
The population of the mobile users is growing at a rapid rate and that the applications
are becoming more bandwidth intensive, hence an efficient resource management and
the traffic control of mobile networks has become a necessary element of mobile communications.
Due to the limitations of the radio spectrum, future wireless systems will
use micro/pico-cellular architectures in order to provide a higher capacity. Because
of the small coverage area of micro/pico-cells and the characteristics of the multipath
and shadow fading radio environment, hand-off events in future systems will occur at
a much higher rate as compared to today’s macro-cellular systems.
Recent research has shown that the algorithms of Artificial Neural Networks (ANN)
have many computational advantages, like massively parallel computation, ability to
learn and adaptive features. Also, an easy implementation of neural networks in analog
circuits and potential powerful computational model for solving real time and large
scale Linear Programming (LP) problems, will help in real time computation. By
considering the above advantages of ANN, our research work involves in exploring the
applications of neural network computational model for resolving the issues in mobile
networks, such as admission control, Resource allocation, location management, mobile
computing and secure Virtual Private Network (VPN) in mobile networks, etc.
Design of algorithms for efficient routing in the Internet
Internet routing or rapid fluctuation of network reachability information, is an
important problem currently facing the Internet Engineering Community. The stability of
end-to-end Internet path is dependent both on the underlying Telecommunication
switching system as well as Routing Architecture. We have studied the instability problem in the Internet and its effects on the network performance. We have designed a fast self-stabilization for a typical routing protocol. This enables faster node convergence, link convergence and network convergence.
To meet the future routing requirements, we have proposed to design more ﬂexible,quickly adaptable to the changes, reliable, scalable, and rapid stabilizable routing strategies by using intelligent agent technologies. This work is aimed at building the Routing Intelligent Mobile Agents (RIMAs) for the routing in the Internet.
Multimedia Communications Traffic Analysis
We have proposed a new hierarchy of high speed network traffic classification, based on the fact that multimedia sources dealing with pre-stored or pre-orchestrated data generate deterministic traffic. These sources are capable of describing their traffic deterministically at call, burst and cell (or packet) level before connection establishment.
We propose a new mathematical traffic model to represent the traffic generated by a pre-recorded multimedia source based on weighted sum of shifted unit functions. We have also proposed an efficient Call Admission Control (CAC) exclusively meant for deterministic traffic and give a method to integrate this CAC with CAC's meant for statistical traffic. We studied the performance of this CAC by simulating it in an arbitrary topology network considering both infinite buffer and finite buffer availability at each intermediate node in the network.
Design of QoS Algorithms for Multimedia Communication
The technical growth in high speed networks, wireless networks, mobile computing
and powerful end-systems are enabling new types of presentational and conversational
multimedia applications, such as video conferencing, video-on-demand and news-on-demand. The media involved in these applications are temporally continuous in nature.
So, such applications require real time multimedia communication and demand guaranteed
Quality of Service (QoS). QoS of an application gives an intuition of how valuable the
services are. The QoS parameters generally considered in multimedia communication
are: end-to-end delay, packet error rate, delay jitter and bandwidth which are
to be negotiated before launching a presentational or conversational multimedia
In this connection, we have designed couple of techniques.
We have designed a QoS based call admission control scheme for mobile networks. It
In this work we have designed a neural network based linear programming model
which enables to use the resources optimally and advises optimal admission of new applications based on readjusting the bandwidth allocation.
We have designed a mobile agent based QoS management system to satisfy the five
functional principles of QoS architecture, i.e., integration, separation,
transparency, asynchronous resource management and performance. A mobile agent is a
software module that migrates sequentially through a set of nodes, remain stationary
and interact with resources remotely or perform any combination of these three
extremes for reservation of resources. It has proved that mobile paradigm saves
considerable amount of bandwidth, reduces network traffic and enables quicker QoS
We have also designed a QoS adaptation and mixer algorithms for wireless
communication. This work proposes new QoS adaptation procedures to provide desired
QoS in wired/wireless networks which allow the system to recover automatically,
if possible from QoS violations by identifying a new configuration of system units
that might support the initially agreed QoS and by performing a user transparent
transition from the original configuration to the new one. We also propose the
Mixer algorithms that might be used in the case of very bad atmospheric conditions at
the time of unable to support the desired QoS but requires a timely delivery of data.
Design of Multiaccess protocols for Wireless Networks
Optimal sharing of a common transmission medium by multiple users in a computer networking environment has been an area of active research in the recent years. Many multiaccess protocols have been designed, investigated and implemented for a wide range of networks. Most of these protocols have been used in Wireless LANs offers throughput of the order of 0.7 maximum when the propagation delay is less compared to the packet transmission time.
We have designed an optimal channel utilization multiaccess (OCUM) protocol and demonstrated the analytical and simulated results of the protocol throughput performance.
Linnenbank, Venkataram and Mullender have designed a MAC protocol for limited bandwidth wireless networks. The protocol is combination of autonomous algorithms for dynamic channel assignment in a TDMA wireless multimedia system. The throughput per mobile station is higher compared to other MAC protocols, it offers low latency for both real-time and non-real time communication. Furthermore, the throughput and latency remain stable under
Design of knowledge based techniques for Network Management
Communication networks have grown and will continue to develop, becoming more divers, more sophisticated, and consequently more difficulty to operate and maintain. Increased reliability and enhanced flexibility in network management are crucial The assessment, diagnosis, control, planning and design are all tasks necessary to perform effective overall network management.
We have designed an expert system for fault detection and localization in communication networks using a network-specific knowledge and network-independent diagnostic mechanism [Refer Publications].
Bandwidth Reservation Protocol for Virtual Private Networks
A Virtual Private Network(VPN) is a service offering that customers
with private network capabilities, although the service is realized through
the public network resources. The customer perceives a private network based
on logical communication resources.
ATM based Virtual Private Networks (AVPNs) are embedded in public ATM network to provide an efficient transport service
to a multisite customer using the virtual levels of ATM. Venkataram and Verma
have designed a centralized protocol for resource allocation in AVPNs and
compare the performance.
A Toolkit for Implementation Oriented Modeling of Real-time Multimedia Communication A great deal of research has been carried out with the fundamental issues arising in the design and implementation of architecture and
protocols for integrated multimedia networks. A comparative performance
evaluation of these protocols and architecture necessitates theoretically
evaluating and experimentally comparing different architectures, protocols,
topologies and technologies of networks under the same operating
Venkataram, Sandeep and Singha have proposed a toolkit to evaluate the
given multimedia network specifications. The toolkit is based on object
oriented, uses Network Description Language (NDL). It has been developed in
two stages for three years. NDL is an inhouse developed compiler.
A framework of traffic control, Resources reservation, a bandwidth regulation mechanism, and priority schemes have been incorporated in the toolkit. The novelty of the toolkit is that if the existing load and the applications to be scheduled are given as an input it provides the schedule structure, the network resources utilization, and meeting of the quality of service (QOS) of all the schedule applications well in advance.
Communication Protocol Specification and Validation
Computer Communication network and the services
provided by them are becoming more sophisticated, leading to the
proliferation of complex communication protocols. Protocols are
not only required for accessing networks and services, but also
for their efficient operations, including advanced network
management strategies. Thus, there is a growing need for the
development of rigorous methods for the design of communication
protocols and their verifications.
Validation is required to ensure
their correctness and proper operations. There is no single
method that can be conveniently used to model and validate all
protocols. Different classes of protocols require different
specification and validation techniques. There have to be proper
matches between classes of protocols and their specification and
We proposed a logic based method to specify a communication protocol as communicating process, interact with one another using synchronous message exchanges. Based on this protocol specification we have designed some of the protocol validation and termination detector techniques. These investigation proved that validation of very large systems, which have declared has very hard and complicated by time and space requirements, can be solved faster and more accurately. Part of the work is published.
Besides these, the work has been carried out to develop an
optimal conformance test sequence and specification of gateways
The Conformance Testing of Protocols
The conformance testing of protocols is a topic of considerable interest
for meeting the objectives of open system interconnection. The main focus of
the recent research on protocol conformance testing has been to find good
optimization techniques that reduce the length of protocol conformance test
sequence. There have been studies on the quality of the test sequences
generated by some of the optimization techniques. There have also been efforts
in classifying faults and characterization of the faults that are detected
by a certain test method.
Premkumar and Venkataram have designed a method for optimal test sequence
generation using Multiple Unique Input/Output Sequences (MUIOS). The test
sequence problem is viewed as an asymmetric traveling sales person (TSP)
problem and is solved by using Simulated Annealing.
Testing Communication Systems
In the area of testing communication systems, the interfaces between systems
to be tested have great impact on test generation and fault detectability.
A general distributed test architecture, containing distributed interfaces,
has been presented for testing distributed systems based on the Open
Distributing Processing (ODP).
Luo, Venkataram and Bochmann have studied the issue of test selection
with respect to such a test architecture. In particular, communication systems
that can be modeled by finite state machine(FSM) with several distributed
interfaces. A test generation method is developed for generating test
sequences for such FSMs, which is based on the idea of synchronizable test
sequences. They also investigated the issue of fault coverage and point out
a fact that given in the literature for the distributed test architecture
cannot ensure the same fault coverage as the corresponding original testing
Development of an expert system for a process plant diagnosis
In recent years, the increasing complexity of a process
plant (either it is a chemical or electrical) and the demand for
safety have created an urgent need for the development of
automated fault diagnosis. There are basically two approaches to
automated fault diagnosis. First is the experience-oriented
method, which is based on the list processing, the pattern
recognition technique or the knowledge base system. This method
is mainly centered around structural knowledge of the plant. The
other is logic oriented method which utilizing a model of the
system being diagnosed, consists of the inverse operation of the
cause-effect relationship. This is generally named as
of the plant.
Planning to have diagnosis system strictly based on either
behavior of a plant or structure of a plant,
available, have not proved to be very efficient. Mixing
two features: structure and behavior of a plant along
expert feels about certain fault symptoms of the plant
of a great help.
We explored this concept of hybrid knowledge acquisition
for process plant diagnosis. The hybrid knowledge is the integration
of shallow empirical knowledge about the structure,
and function of a process plant. We have incorporated this
hybrid knowledge as a part of an expert system for fault diagnosis
The second part of the expert system; the Inference Engine
(IE) is being developed. A portion of the work has been reported