The ICL distributed Array Processor (DAP) is an SIMD array processor containing a large, 2-D array of bit serial processing elements. The architecture of the DAP makes it well suited to data processingapplications wh...
详细信息
The ICL distributed Array Processor (DAP) is an SIMD array processor containing a large, 2-D array of bit serial processing elements. The architecture of the DAP makes it well suited to data processingapplications where searching operations must be carried out on large numbers of data records. This paper discusses the use of the DAP for two such applications, these being the scanning of serial text files and the clustering of a range of types of database. The processing efficiency of the DAP, when compared with a serial processor, is greatest when fixed length records are processed.
One of the key ingredients in writing self-healing parallelapplications that can be migrated from one resource to another in a heterogeneous computing environment like a grid is checkpointing. Checkpointed applicatio...
详细信息
ISBN:
(纸本)1601320841
One of the key ingredients in writing self-healing parallelapplications that can be migrated from one resource to another in a heterogeneous computing environment like a grid is checkpointing. Checkpointed applications have the potential to be reliable, fault tolerant and flexible and hence are useful in ensuring desired quality of service to the user. Complete reengineering of already existing parallelapplications having millions of lines of code to embed checkpointing logic is a daunting task, mainly due to the time and cost overheads involved in the debugging and reengineering process. Aspect-Oriented Programming (AOP) [1 ] techniques are helpful in non-invasive reengineering of applications and promote code reusability and adaptability. This paper investigates the use of AOP techniques to checkpoint parallelapplications to evaluate the benefits of non-invasive reengineering of legacy parallelapplications using AOP techniques. This paper also demonstrates that a checkpointed parallel application can be moved from one resource to another without compromising on accuracy and without any significant decrease in performance.
We consider the broadcasting problem in the Hypercube and Folded Hypercube assuming the all-port shouting mode of communication in the presence of random and dynamic faults. This problem was examined initially using a...
详细信息
ISBN:
(纸本)1892512416
We consider the broadcasting problem in the Hypercube and Folded Hypercube assuming the all-port shouting mode of communication in the presence of random and dynamic faults. This problem was examined initially using a combinatorial approach;however, problems with this approach led to analysis via a computer simulation. The simulation proves the extreme reliability for the Hypercube and the Folded Hypercube, even in the presence of a high percentage of faults.
To support component-based software engineering, simple and efficient mechanisms for dynamic composition and decomposition of components are needed. Shared data spaces are a simple composition mechanism, yet their eff...
详细信息
ISBN:
(纸本)1892512459
To support component-based software engineering, simple and efficient mechanisms for dynamic composition and decomposition of components are needed. Shared data spaces are a simple composition mechanism, yet their efficient distributed implementation faces several complicating factors. One of these factors is that the communication needs of components may differ per data type, per application, and may even change over time. While existing data-space implementations treat all data equally, we propose a distributed data-space architecture that provides the means for differentiating distribution policies according to the type of data. Using this approach we are able to cater for the specific needs of the data. We maintain the transparency of the shared data space paradigm to the application programmer, but extend its capabilities for optimizing its efficiency.
Traditionally network monitoring and management has been done using predominantly centralized techniques. Mobile agents have been proposed as an alternative to this centralized approach. In this paper we propose a nov...
详细信息
ISBN:
(纸本)1892512416
Traditionally network monitoring and management has been done using predominantly centralized techniques. Mobile agents have been proposed as an alternative to this centralized approach. In this paper we propose a novel approach for distributed and dynamic network monitoring, using mobile agents. We use the IBM Aglets system and show how a Java-based distributed network monitoring application can use this paradigm for efficient data collection and analysis and adapt to variations in network characteristics.
The experience in porting NAS NPB micro-benchmark using the recently developed GCC-SGI UPC compiler on an Origin O3800 NUMA machine was reported. The utility of some optimization hints in UPC programming was tested us...
详细信息
ISBN:
(纸本)1892512416
The experience in porting NAS NPB micro-benchmark using the recently developed GCC-SGI UPC compiler on an Origin O3800 NUMA machine was reported. The utility of some optimization hints in UPC programming was tested using the low-level monitoring and specific tuning to UPC code. Thus, a Test Suite which presents a set of tests of I/O performance written using UPC-I/O routines was shown.
Sequence comparison is a task performed in several Bioinformatics applications daily all over the world. Algorithms that retrieve the optimal result have quadratic time complexity, requiring a huge amount of computing...
详细信息
ISBN:
(纸本)9781728165820
Sequence comparison is a task performed in several Bioinformatics applications daily all over the world. Algorithms that retrieve the optimal result have quadratic time complexity, requiring a huge amount of computing power when the sequences compared are long. In order to reduce the execution time, many parallel solutions have been proposed in the literature. Nevertheless, depending on the sizes of the sequences, even those parallel solutions take hours or days to complete. Pruning techniques can significantly improve the performance of the parallel solutions and a few approaches have been proposed to provide pruning capabilities for sequence comparison applications. This paper proposes and evaluates a variant of the block pruning approach that runs in multiple GPUs, in homogeneous or heterogeneous environments. Experimental results obtained with DNA sequences in two testbeds show that significant performance gains are obtained with pruning, compared to its non-pruning counterpart, achieving the impressive performance of 694.8 GCUPS (Billions of Cells Updated per Second) for four GPUs.
The study of distributed systems is increasingly fundamental to a Computer Science curriculum. Yet, the design of applications to run over distributed systems is complex and mastery of fundamental concepts is challeng...
详细信息
ISBN:
(纸本)1892512416
The study of distributed systems is increasingly fundamental to a Computer Science curriculum. Yet, the design of applications to run over distributed systems is complex and mastery of fundamental concepts is challenging for students. In order to assist in distributed systems instruction, we have developed ConcurrentMentor, a visualization system for distributed programming. This system reveals the behavior of a distributed program and its underlying communication protocols while the program executes. Input to the visualization system is generated by an accompanying communication library that closely follows abstractions of communication found in distributed systems literature. No program instrumentation is required.
The current trend towards the use of computing "Grids" provides for the development of more sophisticated job managers. Each of the currently available job managers bring their own benefits to Grid computing...
详细信息
ISBN:
(纸本)1892512416
The current trend towards the use of computing "Grids" provides for the development of more sophisticated job managers. Each of the currently available job managers bring their own benefits to Grid computing. This paper details our initial investigation into using the WebCom meta-computer to schedule the execution of tasks on a Grid running Globus. The immediate benefits of this approach mainly lie in creating Grid applications that utilise WebCom's advanced fault tolerance and load balancing schemes.
This paper presents a new strategy for sharing global information among nodes in a cluster system working with Message Passing Interface (MPI). This strategy combines different MPI processes for communication between ...
详细信息
ISBN:
(纸本)1892512416
This paper presents a new strategy for sharing global information among nodes in a cluster system working with Message Passing Interface (MPI). This strategy combines different MPI processes for communication between nodes and job execution. In order to illustrate this new strategy, an implementation of The Travelling Salesman Problem and its speedups are presented.
暂无评论