Current Research Interests
I am interested in designing metro-scale wireless networks inexpensively using various wireless technologies. The cellular broadband technology complemented with the emerging WiMAX and WiFi mesh technologies can provide inexpensive and easily reconfigurable network within a city with shared access to the Internet. Ad hoc network routing concepts can be beneficial in using multiple wireless technologies and covering a city without gaps. Ad hoc networks use a standards based wireless technology such as Wi-Fi for communication links among energy-constrained, mobile user devices. Owing to node movement and radio interference, communication links are frequently broken temporarily or permanently. So a routing protocol for ad hoc networks must be adaptive and efficient. Traditional proactive routing techniques such as distance vector (DV) and link-state packet (LSP) routing have too much overhead for low bandwidth ad hoc networks. Recently several on-demand routing schemes with low overhead have been proposed. However, such techniques may not handle QoS constraints gracefully. Furthermore, the proactive and on-demand routing techniques represent two extreme points of design. While there have been proposals to mix these two techniques such that parts of a network use DV or LSP and the rest of the network uses on-demand, we believe that routing protocols that are partly pro-active but not overly so with on-demand techniques as the fall back techniques might be beneficial. I have proposed one such routing protocol, denoted adaptive distance vector (ADV), which is shown to work well under heavy traffic load.
Another problem with ad hoc networks is the security. Since routing information is disseminated freely, and since network-wide flooding is used to search for routes, it is easy to launch leveraged, stealth attacks when on-demand routing protocols are used. We are investigating the impact of low-rate denial of service attacks and false route propagations by a small number of malicious nodes on ad hoc networks. My students and I developed a statistical profiling technique to identify dos attackers and mitigate their impact. We have implemented this technique on testbed of 8 Linksys G routers reprogrammed with wrt Linux and AODV software. In both simulations and in the testbed, the attacker is identified quickly. We are using this technique to identify other types of attack such as false route propagation.
Another problem we are addressing is how to maintain network throughput in severely congested networks. My students and I developed several MAC and routing protocol solutions that are easy to implement without impacting the compatibility with the existing protocols. These techniques provide significant performance benefit when the network is congested.
Common standards such as TCP/IP, HTTP, and SNMP and killer applications such as e-commerce, voice on IP ensure further explosive growth and increased importance of the Internet in computing and communications. Furthermore, wireless and ad hoc networks will form a bigger share of the future Internet, LANs and WANs. To provide seamless integration of diverse network technologies, the existing network and application protocols need to be tweaked and new protocols need to be developed. I am working on the performance analysis and enhancement of TCP protocols for multihop, wireless networks.
For my Ph.D. dissertation, I have worked on self-routing algorithms for multistage networks and hypercubes and data alignment and access techniques. The self-routing algorithms can be used to route the class of affine linear permutations, which includes commonly used bit reversal and transpose permutations and many more permutations used in previously proposed data storage and access methods. Based on this observation, I developed data storage methods that use Omega network and provide network and memory conflict-free access to various templates of a stored data array. Significant publications based on this work include papers in JPDC 1991, JPDC 1995, and IEEE TC 1991.
After joining UTSA in 1991, I initiated work on the design of wormhole-switched multicomputers networks and routing algorithms. The main results are design of adaptive routing algorithms compatible the with buffered wormhole switching used in IBM's SP-2 network design, fault-tolerant routing techniques for mesh networks to handle faults on-the-fly with local fault information, and multicast routing methods.
The new adaptive algorithms proposed work for any network topology. It is shown that these algorithms give better throughputs and can be implemented with less buffer space than previously proposed adaptive wormhole routing algorithms. Significant publications based on this work include papers in IEEE TPDS '96 and ISCA '93.
The multicast routing work analyzed a new form of deadlocks that occur with multicast communication in wormhole networks and proposed techniques to remedy the same. Also, the commonly used e-cube routing is adapted for multicast routing and is shown using extensive simulation studies that its performance is superior to many known and contemporary multicast techniques. This work was published in papers in IEEE TPDS '98 and SPDP '94.
The work on fault-tolerant routing proposed the concept of fault rings which contain faults in a network and use the fault-rings to provide on-the-fly alternate routing of messages that are otherwise blocked by faults when the original routing is used. This work has been applied to a variety of networks including to the Cray T3D network and various routing algorithms. Since then this technique has been studied and referenced frequently by other researchers in this area. Significant publications on this work include papers in IEEE TC '95, IEEE TC '97, IEE Proceedings '95, Supercomputing '94 and HPCA '96.
Because of increasing electronic commerce and voice transmissions on the Internet, fast switches that can sustain multi-gigabit and terabit throughputs will be needed. These switches will need to support fast packet switching, stringent quality of service (QoS) guarantees, multicasts, and common standards. Furthermore, fault-tolerance capabilities will become even more crucial as more and more business is transacted on the Internet. I am currently investigating high-speed switch designs that can be scaled from gigabits/sec to terabits/sec throughputs and support a variety of traffic patterns (best-effort, QoS, multicasts). I am applying some of my earlier fault-tolerant routing techniques to handle faults on-the-fly with local information, while avoiding major network outages. A couple of results in this direction were presented at ICC '99 and HiPC '03.
Rajendra V. Boppana