Catálogo de publicaciones - libros

A flat mobile ad hoc network has an inherent scalability limitation. When the network size increases, per node throughput of an ad hoc network rapidly decreases. This is due to the act that in large scale networks, flat structure of networks results in long hop paths which are prone to breaks. These long hop paths can be avoided by building a physically hierarchical backbone network. These networks have some specific backbone capable nodes that have powerful radios and are functionally more capable than ordinary nodes. In this paper, a hybrid routing protocol for large scale networks with mobile backbones has been proposed. This routing protocol uses different types of routing schemes in different layers of hierarchical network which makes it easily extendable to support QoS as well. Along with hierarchical structure, a low-overhead clustering scheme to elect backbone nodes has been proposed and works with our routing protocol without causing any extra overhead.

In this paper, we address the problem of video multicast over ad hoc wireless networks. Video multicasting demands high quality of service with a continuous reachability to receivers. However, the existing multicast solutions do not guarantee this because they are not resilient to mobility of the nodes. Uninterrupted video transmission requires continuous reachability to receivers which emphasizes the usage of path-diversity. Hence, we propose a multiple tree multicast protocol ( K -Tree) which maintains maximally node-disjoint multicast trees in the network to attain robustness against path breaks. We further enhance the robustness by using the error resilient Multiple Description Coding (MDC) for video encoding. Through simulations we show how the protocol improves the video quality as we use two or three trees instead of a single tree.

Ad hoc wireless networks possess highly constrained energy resources. Even the energy aware protocols for ad hoc networks do not consider all the characteristics of the underlying batteries. Hence, they fail to efficiently utilize the available energy. Thus, a mechanism is required to measure the efficiency of the protocols of ad hoc networks, in terms of the network lifetime. To the best of our knowledge, there has been no reported work till date for analyzing the lifetime of the ad hoc networks for various protocols. This paper primarily provides a novel generalized analytical model for estimating lifetime of ad hoc networks, in the presence of the following two kinds of MAC protocols: (i) reservation-based TDMA protocols and (ii) a specific class of CSMA protocols that try to follow a pattern, such as a round-robin scheduling, for packet transmission. We prove through analytical and simulation studies that energy awareness is crucial in deciding the performance of the MAC protocols.

Mobile Grid is an emerging and prospering field of distributed computing where mobile devices are enjoying the benefits of Grid. Challenges faced by mobile Grid are unpredictable network quality, lower trust, limited resources (battery power, network bandwidth, storage, processing power, etc) and extended periods of disconnections which may result in lost of the work done by these devices. We, therefore, need a proper fault tolerance scheme for these mobile hosts. A major issue is the appropriate handling of failures with minimal processing and storage overhead on mobile hosts. To meet these goals, we propose a proxy-based coordinated checkpointing scheme for our mobile to Grid middleware, Mobile Access to Grid Infrastructure (MAGi). In this scheme mobile hosts seamlessly store checkpoints on their respective proxies running on the middleware. Together with the central coordinator component, these proxies act as a centralized checkpointing store. This approach makes it efficient to rollback to the latest consistent global snapshot, without direct involvement of the mobile hosts, which results in less processing and storage overhead on mobile device as compared to existing schemes.

It is increasingly becoming evident that operating system interference in the form of daemon activity and interrupts contribute significantly to performance degradation of parallel applications in large clusters. An earlier theoretical study has evaluated the impact of system noise on application performance for different noise distributions [1]. Our work complements the theoretical analysis by presenting an empirical study of noise in production clusters. We designed a parallel benchmark that was used on large clusters at SanDeigo Supercomputing Center for collecting noise related data. This data was fed to a simulator that predicts the performance of collective operations using the model of [1]. We report our comparison of the predicted and the observed performance. Additionally, the tools developed in the process have been instrumental in identifying anomalous nodes that could potentially be affecting application performance if undetected.

Information-sharing is a key aspect of distributed applications such as database servers and web servers. Information-sharing also assists services such as caching, reconfiguration, etc. In the past, information-sharing has been implemented using ad-hoc messaging protocols which often incur high overheads and are not very scalable. This paper presents a new design for a scalable and a low-overhead Distributed Data Sharing Substrate (DDSS). DDSS is designed to support efficient data management and coherence models by leveraging the features of modern interconnects. It is implemented over the OpenFabrics interface and portable across multiple interconnects including iWARP-capable networks in LAN/WAN environments. Experimental evaluations with networks like InfiniBand and iWARP-capable Ammasso through data-center services show an order of magnitude performance improvement and the load resilient nature of the substrate. Application-level evaluations with Distributed STORM achieves close to 19% performance improvement over traditional implementation, while evaluations with check-pointing application suggest that DDSS is highly scalable.

Failures are likely to be more frequent in systems with thousands of processors. Therefore, schemes for dealing with faults become increasingly important. In this paper, we present a fault tolerance solution for parallel applications that proactively migrates execution from processors where failure is imminent. Our approach assumes that some failures are predictable, and leverages the features in current hardware devices supporting early indication of faults. We use the concepts of processor virtualization and dynamic task migration, provided by Charm++ and Adaptive MPI (AMPI), to implement a mechanism that migrates tasks away from processors which are expected to fail. To demonstrate the feasibility of our approach, we present performance data from experiments with existing MPI applications. Our results show that proactive task migration is an effective technique to tolerate faults in MPI applications.

Clustered architecture processors are preferred for embedded systems because centralized register file architectures scale poorly in terms of clock rate, chip area, and power consumption. Although clustering helps by improving clock speed, reducing energy consumption of the logic, and making the design simpler, it introduces extra overheads by way of inter-cluster communication. This communication happens over long global wires which leads to delay in execution and significantly high energy consumption. In this paper, we propose a new instruction scheduling algorithm that exploits scheduling slacks of instructions and communication slacks of data values together to achieve better energy-performance trade-offs for clustered architectures with heterogeneous interconnect. Our instruction scheduling algorithm achieves 35% and 40% reduction in communication energy, whereas the overall energy-delay product improves by 4.5% and 6.5% respectively for 2 cluster and 4 cluster machines with marginal increase (1.6% and 1.1%) in execution time. Our test bed uses the Trimaran compiler infrastructure.

Performance of large memory applications degrades rapidly once the system hits the physical memory limit and starts paging to local disk. We present the design, implementation and evaluation of Distributed Anemone (Adaptive Network Memory Engine) – a lightweight and distributed system that pools together the collective memory resources of multiple machines across a gigabit Ethernet LAN. Anemone treats remote memory as another level in the memory hierarchy between very fast local memory and very slow local disks. Anemone enables applications to access potentially “unlimited” network memory without any application or operating system modifications. Our kernel-level prototype features fully distributed resource management, low-latency paging, resource discovery, load balancing, soft-state refresh, and support for ’jumbo’ Ethernet frames. Anemone achieves low page-fault latencies of 160 μs average, application speedups of up to 4 times for single process and up to 14 times for multiple concurrent processes, when compared against disk-based paging.

Localization, an important challenge in wireless sensor networks, is the process of sensor nodes self-determining their position. The difficulty encountered is in cost-effectively providing acceptable accuracy in localization. The potential for the deployment of high density networks in the near future makes scalability a critical issue in localization. In this paper we propose Cluster-based Localization (CBL), which provides effective localization suitable for large and highly-dense networks. CBL utilizes both a computationally-intensive localization technique (non-metric multidimensional scaling (MDS)) and a less intensive trilateration to achieve balance between performance and cost. Clustering is utilized to select a subset of nodes to perform MDS and then extend their localization to the remaining network. Besides providing scalability clustering overcomes local irregularities and provides good accuracy even in irregular networks with or without obstacles. Simulation results illustrate that CBL reduces both computation and communication, while still yielding acceptable accuracy.