in this paper we discuss the peer to peer solution for the deployment of large scale numerical applications on distributed resources. We use the XtremWeb P2P middleware for managing the computational resources. this g...
详细信息
ISBN:
(纸本)0769522564
in this paper we discuss the peer to peer solution for the deployment of large scale numerical applications on distributed resources. We use the XtremWeb P2P middleware for managing the computational resources. this global computing platform, such as much current grid deployments, is currently dedicated to multi-parameters applications. our goal is to use it for executing numerical parallel applications that require communications between tasks. We discuss the performances of a block based matrix-vector product deployed for two XtremWeb networks;a local one with 128 non dedicated PCs, and for 261 PCs distributedon two geographic sites: University of Lille i and Paris Xi University at orsay. We give an investigation of the scheduling schemes well adapted for such systems and applications. We also present the performances of out-of-core products and discuss relationship between out-of-core and peer-to-peer computing for numerical applications.
this paper presents a natural algebraic specification for the P systems. the specification is executable in Maude, a software system supporting rewriting and equational logic. We define the P system maximal parallel e...
详细信息
ISBN:
(纸本)3540250808
this paper presents a natural algebraic specification for the P systems. the specification is executable in Maude, a software system supporting rewriting and equational logic. We define the P system maximal parallel evolution as a specific rewriting strategy in Maude. By extending the Maude rewriting semantics withthis strategy,We provide an operational semantics of the P systems. We present few examples of specifying and executing simple P systems, describing how target indications, dissolving and priorities are handled. Moreover, the Maude system allows the verification of various properties of the P systems expressed as linear temporal logic formulas by using a model checker.
distributed Shared i/o System is an important research field of multiprocessor system. HyperTransport is an emerging interconnection standard that offers high-speed, high-performance, point-to-point packet-based link ...
详细信息
ISBN:
(纸本)3540200541
distributed Shared i/o System is an important research field of multiprocessor system. HyperTransport is an emerging interconnection standard that offers high-speed, high-performance, point-to-point packet-based link between integrated circuits on a board, particularly between a high-performance processor and peripheral devices. in this paper we introduce a new high-performance distributed shared i/o system developed by ourselves, which uses HyperTransport technology toimprove i/o performance and scalability. it attempts to extend the advantages of SMP withthe benefits of MPPs and clusters, and overcome the distant i/o problem of distributed non-shared i/o system. To concentrate on a study of the i/o architectural design, some key technologies implemented in the system are discussed.
the proceedings contain 77 papers. the special focus in this conference is on Architecture, Software, Grid and Network. the topics include: Using split queues toimprove the performance of parallel switch;a data drive...
ISBN:
(纸本)3540200541
the proceedings contain 77 papers. the special focus in this conference is on Architecture, Software, Grid and Network. the topics include: Using split queues toimprove the performance of parallel switch;a data driven loop engine on array processor;a new architecture of a fast floating-point multiplier;a highly efficient FC-san based on load stream;a new high-performance distributed shared i/o system;design and implementation of Fortran front end;an alternative superscalar architecture withinteger execution units only;a high efficiency distributed mutual exclusion algorithm;the security architecture of the java operating system JX;simultaneous multithreading trace processors;a VLSi architecture design of 1-d dwt;overcoming static register pressure for software pipelining in the itanium architecture;separating data storage, data computation, and resource management one from another in operating systems;a high performance design and implementation of the virtual interface architecture;optimization of asynchronous volume replication protocol;predicate analysis based on pathinformation;new no-blocking permutation network;a task-oriented parallel computation model;a method of data assignment on heterogeneous disk system;orthogonal design method for optimal cache configuration;a probabilistically correct election protocol in asynchronous distributedsystems;a formal specification and method for MAS as a distributed system;optimal fixed priority assignment with limited priority levels;a proof assistant for mobile processes and data space fusion based approach for effective alignment of computation and data.
one of the main problems arising when designing distributed High Performance programs, or when restructuring sequential programs for distributedsystems, is workload unbalance. the Client-Server paradigm, which is usu...
详细信息
ISBN:
(纸本)354041729X
one of the main problems arising when designing distributed High Performance programs, or when restructuring sequential programs for distributedsystems, is workload unbalance. the Client-Server paradigm, which is usually adopted for distributedsystems, is not flexible enough toimplement effective dynamical workload balancing strategies. the Mobile Agent paradigm can increase the flexibility in the creation of distributed applications (and the restructuring of sequential applications for distributedsystems), and can in particular provide with a robust framework for managing dynamical workload balancing. in this paper we show, through a case-study, how the restructuring of a sequential code implementing an irregular algorithm, with adoption of the mobile agent model, allows for yielding a load-balanced distributed version of the algorithm without completely rethinking its structure, and for reusing a great deal of the sequential code.
the proceedings contains 10 papers. Topics discussed include distributedsystems, prefetching and caching, out-of-core issues and disk arrays and input/output interconnects.
the proceedings contains 10 papers. Topics discussed include distributedsystems, prefetching and caching, out-of-core issues and disk arrays and input/output interconnects.
the proceedings contain 22 papers. the special focus in this conference is on Static Analysis and openMP Support. the topics include: A collective i/o scheme based on compiler analysis;achieving robust, scalable clust...
ISBN:
(纸本)3540411852
the proceedings contain 22 papers. the special focus in this conference is on Static Analysis and openMP Support. the topics include: A collective i/o scheme based on compiler analysis;achieving robust, scalable cluster i/oin java;high level programming methodologies for data intensive computations;static analysis for guarded code;a framework for efficient register allocation through selective register demotion;a comparison of locality transformations for irregular codes;a runtime system for tuning the memory performance of openMP programs on scalable shared-memory multiprocessors;performance evaluation of openMP applications with nested parallelism;adaptive parallelism for openMP task parallel programs;optimizing mutual exclusion synchronization in explicitly parallel programs;detecting read-only methods in java;the effect of contention on the scalability of page-based software shared memory systems;measuring consistency costs for distributed shared data;compilation and runtime optimizations for software distributed shared memory;run-time support for distributed sharing in typed languages;a middleware system for distributed shared state;run-time support for adaptive heavyweight services;an infrastructure for monitoring and management in computational grids;realistic CPU workloads through host load trace playback;thread migration and load-balancing in heterogenous environments;toward compiler support for scalable parallelism using multipartitioning and speculative parallelization of partially parallel loops.
out-of-core applications perform poorly in paged virtual memory (VM) systems because demand paging involves slow disk i/o accesses. Much research has been done on reducing the i/ooverhead in such applications by eith...
详细信息
out-of-core applications perform poorly in paged virtual memory (VM) systems because demand paging involves slow disk i/o accesses. Much research has been done on reducing the i/ooverhead in such applications by either reducing the number of i/os or lowering the cost of each i/ooperation. in this paper, we investigate a method that combines fine-grained threading with a memory server model toimprove the performance of out-of-core applications on multicomputers. the memory server model decreases the average cost of i/ooperations by paging to remote memory, while the fine-grained thread scheduling reduces the number of i/o accesses by improving the data locality of applications. We have evaluated this method on an intel Paragon with 7 applications. our results show that the memory server system performs better than the VM disk paging by a factor of 5 for sequential applications and a factor of 1.5 to 2.2 for parallel applications. the fine-grained threading alone improves the VM disk paging performance by a factor of 10 and 1.2 to 3 respectively for sequential and parallel applications. overall, the combination of these two techniques outperforms the VM disk paging by more than a factor of 12 for sequential applications and a factor of 3 to 6 for parallel applications.
the proceedings contain 41 papers. the topics discussed include: large-grain, dynamic control system architectures;multi-paradigm programming in large control systems;use cases for distributed real-time software archi...
ISBN:
(纸本)0818680962
the proceedings contain 41 papers. the topics discussed include: large-grain, dynamic control system architectures;multi-paradigm programming in large control systems;use cases for distributed real-time software architectures;toward a scalable design for command and control systems;integrated scheduling of tasks and messages in distributed real-time systems;prioritized token-based mutual exclusion for distributedsystems;synchronization for distributed real-time applications;distributed real-time problem statement: the air defense system;and comparing two-phase locking and optimistic concurrency control protocols in multiprocessor real-time databases.
暂无评论