Browsing by Author "Dr. Rudra Dutta, Committee Chair"

Now showing 1 - 10 of 10

Design and Implementation of a Distributed Scheduling Algorithm using Period Inflation for Sensor Networks.
(2007-03-22) Bhatt, Bhushan; Dr. Rudra Dutta, Committee Chair; Dr. Do Young Eun, Committee Member; Dr. Mihail Sichitiu, Committee Member
Wireless Sensor Networks (WSNs) are fast emerging as a new and ubiquitous networking arena which will enables many new applications and pervades many old ones. One of the motivations for the development of WSNs is their ability to be deployed in any environment in a comparatively ad-hoc manner. The most important challenge faced by WSNs is battery-limited lifetime of the network. Physically replacing batteries is infeasible in most real-life deployments of WSNs. It has been demonstrated both theoretically and practically that intelligent operation of WSN nodes can improve network lifetime. For example, turning off wireless transceivers at WSN nodes, minimizing idle listening, can increase battery lifetimes by large factors, especially in many passive data sensing applications where the sense-receive-transmit cycle of the sensors is periodic. In particular, we focus on some previous work in which an adaptive scheduling algorithm was proposed for this purpose, under unpredictable but small clock drift (so called quasi-periodic traffic). While this approach can adapt effectively to unknown transmission periods and unknown changes in transmission periods, the fundamental problem remains: a few nodes close to the base station deplete their batteries sooner than the rest resulting in early network death. Further, this phenomenon reduces the effectiveness of the method even more when (a) the periods of the various nodes are very disparate, and (b) when nodes artificially reduce their periods to maintain end-to-end delay bounds. In this thesis, we advance a new technique called "period inflation", by which the nodes of a WSN can cooperatively create a schedule in which nodes close to the base station have higher periods. We investigate the performance of the inflated and non-inflated cases for scenarios where all nodes have similar periods as well as when some nodes have very disparate periods, and also under bounded delay conditions. Numerical results show that the new technique of period inflation performs better, as expected.
Joint Scheduling, Routing and Power Control for Single-Channel Wireless Mesh Networks
(2006-08-29) Gupta, Divya; Dr. Mihail Sichitiu, Committee Member; Dr. Harry Perros, Committee Member; Dr. Rudra Dutta, Committee Chair
Mesh networks is a class of wireless networks consisting of a set of backbone nodes and some client nodes. In this work, we look at the problems of routing, scheduling and power control in such networks, with the ultimate goal of increasing the throughput while satisfying all the traffic requirements. Increasing the throughput of the network implies the ability to send more data between as many source and destination nodes as possible, in shorter periods of time. To achieve this, we need intelligent scheduling schemes that take advantage of the spatial reuse that is possible. We outline two heuristic scheduling algorithms that look at various ways of ordering the links and choosing the ones that might reduce the schedule length, which is a measure of the throughput of the network. The simpler algorithm works at the candidate node level for each slot, while the more complex independent sets based algorithm works with individual links. Different options have been outlined for choosing the first and subsequent links of every slot in the schedule. This includes the interference score of the link, the magnitude of traffic requirement, the specific interfering links, etc. We also study in detail how the links in the network affect the schedule length. When certain links are scheduled, the spatial reuse factor is reduced to zero, implying that no other links can be active at the same time as these links. Hence it is of interest to us to study more about these "loner" links as they add to the schedule length. We characterize these links for topologies in square and circular areas. Using simple geometric arguments, we show that links of length 0.579k and 0.485d in a square area of side k and circular area of diameter d respectively, will always be loners. We also outline a method to analytically find the number of loner links in a given network. This number gives a lower bound on the schedule length. With the understanding gained from the study of scheduling heuristics, we present a routing algorithm that performs topology control by removing loner links. The traffic on these links is routed using other links in the network. Although this routing scheme does not perform well, it leads to a more intelligent joint routing and scheduling mechanism that succeeds in reducing schedule lengths. The joint mechanism does routing in parallel with 2 the scheduling, by calculating routes over links that fit completely into the current schedule. The power control mechanism for both these schemes is to use the minimum power level needed for communication.
Network Design and Optimization - with Applications in Optical and Wireless Networks
(2006-09-07) Huang, Shu; Dr. Carla Savage, Committee Member; Dr. David Thuente, Committee Member; Dr. Harry Perros, Committee Member; Dr. Rudra Dutta, Committee Chair
Typically, enterprise networks may incorporate Wide Area Networks (WANs),Metro Area Networks (MANs) and Local Area Networks (LANs). These networks have very different characteristics in terms of the physical media, data link layer and network layer protocols they use. In all these networks, well-designed network architectures are the key to achieving high performance at reasonable costs. We study the optimization problems that have arisen in the design of different networks, specifically, optical networks and wireless networks. In optical networks, we study the traffic grooming problem in Wavelength Division Multiplexing (WDM) networks with dynamic traffic demands. We present of detailed study of current research in this area and propose a new design problem for both single-link and multi-link networks. In wireless networks, we present new formulations for the design problem in Wireless Mesh Networks (WMN) that take different interference models into consideration and propose algorithmic methods to solve them.
Optimization of the Processing in a Middleware Environment for a Set of XML Variants
(2007-03-22) Sahijwani, Vineet; Dr. Rudra Dutta, Committee Chair; Dr. Matthias F. Stallmann, Committee Member; Dr. Rada Chirkova, Committee Member
The middleware has become an important product category since the 1990s, when companies started to adopt client-server architectures and had to deal with distributed computing problems. It employs the component based design to distribute the processing across the network. The middleware components use proprietary formats or more popularly XML for data exchange. There are different formats or variants of XML available which when used can improve the processing speed at a middleware component. However it is possible that the same format may not be suitable or be the best for all the components the XML goes through. Thus from a given set of XML variants or formats, there can be a particular XML format that will process faster as compared to others in a given middleware component. That same XML format however may not be suitable for other middleware components and some other format from the set might be. In this thesis we consider the problem of improving the processing speed in the middleware, given a set of different XML formats. We propose the use of local optimization to achieve global optimization. We find the best suited XML format for each component. This is done with the help of a feedback based software monitoring system. The software collects data about the processing time taken in a particular component by different formats with the help of feedback from that component and then process that data to decide which format suits that component best. Once the formats are known, the data can be sent to the components in that format and it would result in faster processing in the system as a whole, as compared to, if a single format was used for all the components. We also numerically investigate the performance of a middleware environment with and without the feedback based software monitoring system.
Power Conservation in wireless sensor network using receiver switch-off
(2004-03-28) Goel, Apurva Ajay; Dr. David Thuente, Committee Member; Dr. Rudra Dutta, Committee Chair; Dr. Mihail Sichitiu, Committee Member
A wireless sensor network is a network of nodes equipped with sensors and capable of relaying their data to the monitoring station using multi-hop communication. Essentially, they operate using the ad hoc paradigm due to the unpredictable and dynamic topology. The lifetime of the nodes is limited to that of its battery. Thus in this work we try to conserve the battery by putting the nodes to sleep when they are not involved in any communication process. Sensor nodes produce readings every fixed time interval and we try to leverage the predictability of the packet inter-arrival time to obtain the sleep periods. We model the inter-arrival times that a node observes in the form of a probability distribution function. A node may be forwarding packets from multiple sources and it associate this knowledge to every stream it is carrying. After a node captures a packet, using its knowledge it can predict the approximate arrival time of the next packet. It can try to sleep during this period. It thus conserves power. In our study we show that simply forwarding the packet when they are received causes the predictability needed for this approach to be lost. We then propose an alternative approach, which is to maintain some local periodicity. This can be accomplished by delaying the packets. The multi-hop nature of the network causes these delays to accumulate and thus we have to consider the tolerances of the data packets to these delays. We then discuss the relation between the sleep durations possible and the delay tolerances of the packets. Finally, we explore the inter-dependence between packet loss, sleep duration, delay and the ultimate effect on the power savings.
Power-aware Routing: Grooming Traffic for Reduction of Power Consumption in Wire-line Networks
(2008-08-19) Seshadri, Deepa; Dr. Rudra Dutta, Committee Chair; Dr. Douglas Reeves, Committee Member; Dr. Subhashish Bhattacharya, Committee Member
The Internet has experienced an exponential growth with an increase in the number of end users and network connections. The increased demand has been met by increased bandwidth. However, this tremendous growth has also implied that there is an increase in the power consumption in the networks and routers. Additionally, a lot of heat is generated by the routers and there is an increasing probability that heat management issues become the bottleneck in the operation of network. This work aims to address the issue of power-management in wire-line networks by creating an effective routing (called power-aware routing) that would minimize the total number of network interfaces switched on in the routers. The problem of power-aware routing is formulated as an Integer Linear Program. However, as the formulation is found to be computationally prohibitive for networks with large number of nodes, heuristic algorithms are proposed. A path-based heuristic formulation is first proposed. In the path-based algorithm, traffic is routed along the shortest Dijkstra path available over the set of minimum possible 'on' interfaces. While the number of interfaces turned on in this formulation is close to the results from the ILP formulation, the aggressive nature in routing over a minimum possible 'on' interfaces means that the approach is not successful for cases when the traffic load in the network is high. Thus, to obtain solutions for larger range of networks, a second heuristic algorithm, called link-based algorithm, is proposed. In the link-based algorithm, the interfaces are switched on as pairs. The interface-pairs are switched on in the increasing order of their combined power consumption. Although, the number of interfaces that are switched on is higher than what is obtained from the path-based formulation, the link-based heuristic solves for more number of network scenarios. Flow deviation approach is then applied to the solution obtained from the link-based heuristic to provide a tighter upper bound on the number of interfaces required. The proposed heuristic algorithms are compared with the results from the ILP formulation for eight node and ten node networks. Additionally, these algorithms are also compared against the usual routing methods that are in use today. These comparisons show that by employing power-aware routing, considerable amount of power can be saved as against the usual routing methods that are in use today. The number of interfaces that are turned on by the path-based heuristic algorithm is almost one-third of the interfaces turned on by the basic routing algorithm used in networks (Dijkstra's shortest path routing). The link-based heuristic also performs better than basic routing by switching on only half the number of interfaces than that are switched on by basic routing. The heuristic algorithms were also simulated and compared for node numbers greater than 10.
Protection in survivable WDM grooming network.
(2004-05-05) Tiyachate, Ruarob; Dr. Rudra Dutta, Committee Chair; Dr. Mladen A. Vouk, Committee Member; Dr. Injong Rhee, Committee Member
Optical networks have been widely expected to fill the need for tomorrow's backbone networks because of the high bandwidth and highly predictable performance they promise. However, in recent years, the downturn of the economy has made the deployment of costly equipment to obtain very high bandwidth less immediately attractive. In the current context, research must address these realistic conditions, and this is part of the motivation for the area in which the research in this thesis is performed. We propose a two-step approach to design, and show how this approach is suitable from both the grooming and the protection points of view. We adopt well-known heuristics from literature to perform stand-alone grooming and stand-alone protection at the virtual link level, enhancing the protection algorithm by adotping a failure independent routing but failure-specific wavelength assignment for protection virtual links. We show that in relative terms, the grooming performance of the protection design is already quite good due to this approach. However, protecting at the virtual link level invariably increases the grooming cost of the protection solution, and this cost can be significant at the node at which it is maximum. We go on to show how this can be countered by performing sub-wavelength protection. All our theoretical expectations are validated by numerical simulations. Our results are as follows. Subwavelength protection can effectively use unutilized capacity in existing virtual topology. Proposed protection results in protection solution with less total amount of electronic processing from both individual traffic component and overall traffic component point of view. Proposed algorithm not only decreases the amount of electronic processing from protection solution, but also decreases amount of virtual link setup for protection. The implementation of the two-phase approach help reduce the computation time to find an effective protection solution. Numerical results show that proposed algorithm has performed well and produces similar effects on different traffic patterns and topologies.
Reconfiguration of Sub-Wavelength Groomed Wavelength Routed Optical Networks
(2004-02-23) Mahalati, Ruhiyyih; Dr. Rudra Dutta, Committee Chair
Telecommunication networks recently have seen a large increase in traffic demands, especially data traffic as compared to voice traffic. With the advances in fiber optics and wavelength division multiplexing (WDM), optical networking is the key to satisfy the data-driven bandwidth demand. These technologies enable simultaneous transmission of signals on separate high-speed channels at different wavelengths. While current technologies can provide such huge bandwidth, in order to utilize efficiently the capacity of each lightpath, a number of independent lower-rate traffic streams must be multiplexed into a single lightpath. This technique is referred to as traffic grooming. Another most attractive feature of WDM and wavelength routing networks is the possibility of adaptively creating virtual topologies, or a set of lightpaths, based on network need, giving rise to the concept of reconfiguration. Till date, however, there has been little or no work on the joint consideration of the two areas of traffic grooming and reconfiguration, even though it is clear that reconfiguration is at least equally important in the realistic networks of tomorrow which will definitely need to carry sub-wavelength traffic. This is probably due to two reasons: the common wisdom has been that the two aspects can be handled separately, and also it is hard to define reasonable network design goals if the two aspects are considered jointly. In this thesis, we examine this issue of reconfiguration in groomed networks. The overall focus of this work has been the balancing of the reconfiguration cost and a good grooming solution. We define suitable goals for an integrated approach, and provide formulation of the integrated approach as an integer linear program. To allow a joint consideration of grooming and reconfiguration costs, we mathematically formulate a representation of the reconfiguration cost in terms of the OXC and DXC reconfiguration that can be related to grooming costs. We also develop a heuristic strategy for addressing the problem, which attempts to achieve minimal disturbance reconfiguration by performing local reconfiguration and delaying the need for global reconfiguration. In order to proactively avoid global reconfiguration, we introduce the concept of over provisioning at the traffic demand level. Our reconfiguration heuristic minimizes the need for solving the integer linear program, which is computationally intractable. We also present numerical results validating our claims.
Traffic Grooming in Translucent Optical Ring Networks.
(2003-12-07) Srinivasarao, Koundinya Bangalore; Dr. Rudra Dutta, Committee Chair; Dr. George N Rouskas, Committee Member; Dr. Peng Ning, Committee Member
The exponential growth of the Internet has resulted in an ever increasing demand for bandwidth. Carrier networks which form the backbone of the Internet, have been designed to carry only voice signals with predictable traffic patterns and anticipating slow growth of the network. With the advances in fiber optics and wavelength division multiplexing (WDM) optical networking is the key to satisfy the data-driven bandwidth demand. These technologies enable simultaneous transmission of signals on separate high-speed channels at different wavelengths. While the bandwidth provided by these channels is very high, individual traffic demands are at the sub-wavelength level. This mismatch can be overcome by multiplexing several lower rate connections onto the high-speed channels in a cost-effective manner. This technique is referred to as traffic grooming. Traffic grooming in WDM networks has been a widely addressed problem in recent years. Traffic grooming and its constituent subproblems have been proven to be NP-complete for even the most elemental of network topologies. The ring topology has been the target of a large number of the studies because of its practical relevance. However, most existing studies concentrate on some objective function that is aggregated over all the network nodes, such as the total number of ADMs used or the total amount of opto-electro-optical (OEO) routing performed. From a practical point of view, it is likely that every network node would be provisioned similarly. Hence a min-max objective, seeking to minimize the OEO equipment needed at the node which needs the maximum of such equipment is more appropriate. Such objectives are usually harder to optimize than aggregate objectives which are themselves known to be computationally intractable. In this thesis, we study traffic grooming in a unidirectional ring network under different traffic patterns for the min-max objective. We define two heuristic approaches based on decomposition; one is based on grouping the nodes, and the other on partitioning the traffic matrix. We show that the second approach is more general, but is costlier in terms of computation; further, we indicate traffic families for which the first approach may be expected to perform nearly as well as the more complex one. We also investigate several variations of these two main approaches. We present numerical results validating the performance of the algorithms.
Traffic Grooming in Wavelength Routed Path Networks
(2003-01-03) Huang, Shu; Dr. Rudra Dutta, Committee Chair; Dr. G.N.Rouskas, Committee Member; Dr. M.Stallmann, Committee Member
Recent increase in bandwidth and development of wavelength routing techniques has prompted the study of traffic grooming in wide area wavelength routed optical networks. These networks are widely expected to form the high speed high performance backbone networks of the future. We have studied the grooming problem as applied to the path network, where the physical topology of fibers simply forms a path. Path networks are important in themselves, because they arise naturally in different realistic scenarios, but also because they are simple elemental topologies which can contribute to the understanding of more complex topologies. We show that the problem of traffic grooming is NP-Complete in both unidirectional and bidirectional path networks under the objective of minimizing network-wide amount of electronic switching, whether bifurcation of traffic components into integral sub-components is allowed or not. These results are somewhat surprising in such a simple topology as the path, and underline the inherent complexity of the grooming problem. Our results have implications for grooming problems with other topologies, which we explore. We also explore the approximability of the problem in path networks. We prove that there is no such a polynomial approximation algorithm that it can guarantee an approximation ratio less than infinity, unless $P=NP$. We propose a heuristic approach to solve the problem practically. Our approach is loosely based on the idea of flow deviation. After defining the deviation framework in the context of our problem, we show that it is a family of algorithms rather than a single one, the different members of the family obtained by choosing different candidate approaches to two key subtasks. Some of these members possess practically important performance guarantees, which we define. We present numerical results obtained by applying our technique to traffic instances of various patterns to validate our theoretical claims.