Chadi Kari

Data Migration in Heterogeneous Storage Systems

Large-scale storage systems are crucial components in data-intensive applications such as search engine clusters, video-on-demand servers, sensor networks and grid computing. A storage server typically consists of a set of storage devices. In such systems, data layouts may need to be reconfigured over time for load balancing or in the event of system failure/upgrades. It is critical to migrate data to their target locations as quickly as possible to obtain the best performance. Most of the previous results on data migration assume that each storage node can perform only one data transfer at a time. A storage node, however, can typically handle multiple transfers simultaneously and this can reduce the total migration time significantly. Moreover, storage devices tend to have heterogeneous capabilities as devices may be added over time due to storage demand increase. In this paper, we consider the heterogeneous data migration problem, where we assume that each storage node v has different transfer constraint cv, which represents how many simultaneous transfers v can handle. We develop algorithms to minimize the data migration time. We show that it is possible to find an optimal migration schedule when all cvs are even. Furthermore, though the problem is NP-hard in general, we give an efficient algorithm that offers a rigorous (1 + o(1))-approximation guarantee.

The Efficacy of Epidemic Algorithms on Detecting Node Replicas in Wireless Sensor Networks

A node replication attack against a wireless sensor network involves surreptitious efforts by an adversary to insert duplicate sensor nodes into the network while avoiding detection. Due to the lack of tamper-resistant hardware and the low cost of sensor nodes, launching replication attacks takes little effort to carry out. Naturally, detecting these replica nodes is a very important task and has been studied extensively. In this paper, we propose a novel distributed, randomized sensor duplicate detection algorithm called Discard to detect node replicas in group-deployed wireless sensor networks. Our protocol is an epidemic, self-organizing duplicate detection scheme, which exhibits emergent properties. Epidemic schemes have found diverse applications in distributed computing: load balancing, topology management, audio and video streaming, computing aggregate functions, failure detection, network and resource monitoring, to name a few. To the best of our knowledge, our algorithm is the first attempt at exploring the potential of this paradigm to detect replicas in a wireless sensor network. Through analysis and simulation, we show that our scheme achieves robust replica detection with substantially lower communication, computational and storage requirements than prior schemes in the literature.

Randomized Work-Competitive Scheduling for Cooperative Computing on k-partite Task Graphs

A fundamental problem in distributed computing is the problem of cooperatively executing a given set of tasks in a dynamic setting. The challenge is to minimize the total work done and to maintain efficiency in the face of dynamically changing processor connectivity. In this setting, work is defined as the total number of tasks performed (counting multiplicities) by all the processors during the course of the computation. In this scenario, we are given a set of t tasks that must be completed in a distributed setting by a set of p processors where the communication medium is subject to failures. We assume that the t tasks are similar, in that they require the same number of computation steps to finish execution. We further assume that the tasks are idempotent - executing a task multiple times has the same effect as a single execution of the task. The tasks have a dependency relationship defined among them captured by a task dependency graph.

Soft Edge Coloring

We consider the following channel assignment problem arising in wireless networks. We are given a graph G= (V, E), and the number of wireless cards C v for all v, which limit the number of colors that edges incident to vcan use. We also have the total number of channels C G available in the network. For a pair of edges incident to a vertex, they are said to be conflictingif the colors assigned to them are the same. Our goal is to color edges (assign channels) so that the number of conflicts is minimized. We first consider the homogeneous network where C v = kand C G i¾? C v for all nodes v. The problem is NP-hard by a reduction from Edge coloring and we present two combinatorial algorithms for this case. The first algorithm is a distributed greedy method, which gives a solution with at most

Gesture based Human Computer Interaction for Athletic Training

(1 - \frac{1}{k})|E|

Work-Competitive Scheduling on Task Dependency Graphs

more conflicts than the optimal solution. We also present an algorithm yielding at most |V| more conflicts than the optimal solution. The algorithm generalizes Vizings algorithm in the sense that it gives the same result as Vizings algorithm when k= Δ+ 1. Moreover, we show that this approximation result is best possible unless P= NP. For the case where C v = 1 or k, we show that the problem is NP-hard even when C v = 1 or 2, and C G = 2, and present two algorithms. The first algorithm is completely combinatorial and produces a solution with at most

Distributed dynamic channel assignment in wireless networks

(2-\frac{1}{k}) OPT + (1 - \frac{1}{k}) |E|

The strong at-most-once problem

conflicts. We also develop an SDP-based algorithm, producing a solution with at most 1.122 OPT+ 0.122 |E| conflicts for k= 2, and

Epidemic Node Replica Detection in Sensor Networks

(2-\Theta(\frac{\ln k}{k})) OPT + (1 - \Theta(\frac{\ln k}{k}))|E|

A2Cloud: An Analytical Model for Application-to-Cloud Matching to Empower Scientific Computing

conflicts in general.