The continuous increase in the volume of data, decrees the employment of parallel anddistributed Computing. The processing requirements of parallel environments are complex and more stringent than the uniprocessor sy...
详细信息
ISBN:
(纸本)0897916263
The continuous increase in the volume of data, decrees the employment of parallel anddistributed Computing. The processing requirements of parallel environments are complex and more stringent than the uniprocessor systems. Efficient exploitation of parallelism dictates an even partitioning of the computation across the processing sites. To achieve a uniform load, the database optimizers require statistical information of the underlining relations. We developed a query optimization approach, named Dynamic Optimization on Multiprocessor Engines (DOME), that uses a dynamic sampling methodology to determine the frequency distribution along each level of the query tree. DOME covers the three main multiprocessor query optimization areas of Workload Partitioning, Site Selection, and Operation Ordering. We present the Site Selection segment of DOME. The Site Selection segment receives the statistical characteristics of the participating relations from the Workload Partitioning segment and performs three major operations: identifies an allocation scheme between buckets and nodes;partitions the relations residing on the various processor of the distributed environment into the appropriate buckets;transfers the buckets to their corresponding nodes. We implemented DOME on an Intel i860 hypercube with 32 nodes and tested its behavior through experimentation. The Site Selection algorithms provide approximately a six fold factor improvement over a static allocation approach for the Join relational operation.
Building the hardware for a high-performance distributed computer system is a lot easier than building its software. The authors describe a model for programming distributed systems based on abstract data types that c...
详细信息
Multiprocessor systems should exist in the larger context of distributed systems, allowing multiprocessor resources to be shared by those that need them. Unfortunately, typical multiprocessor resource management techn...
详细信息
The proceedings contain 128 papers. The topics discussed include: C parallelizing compiler on local-net work- based computer environment;OCCAM prototyping of massively parallelapplications from colored Petri-nets;per...
ISBN:
(纸本)0818634421
The proceedings contain 128 papers. The topics discussed include: C parallelizing compiler on local-net work- based computer environment;OCCAM prototyping of massively parallelapplications from colored Petri-nets;performance characteristics of the iPSC/SSO and CM-2 I/O systems;automatic parallelization of LINPACK routines on distributed memory parallel processors;transformation of doacross loops on distributed memory systems;an efficient atomic multicast protocol for client-server models;a new horizon for sorting on mesh architectures;mapping of uniform dependence algorithm onto fixed size processor arrays;and towards understanding block partitioning for sparse Cholesky factorization.
Requirements of emerging applications together with rapid changes in networking technology towards gigabit speeds require new adequate transport systems. Integrated designs of transport services, protocol architecture...
详细信息
Requirements of emerging applications together with rapid changes in networking technology towards gigabit speeds require new adequate transport systems. Integrated designs of transport services, protocol architecture, and implementation platforms are needed for the requirements of forthcoming applications in high-speed network environments. The transport subsystem PATROCLOS (parallel Transport subsystem for cell based high speed networks) is designed with special emphasis on a high degree of inherent parallelism to allow efficient implementations on multiprocessor architectures combined with specialised hardware for very time critical functions. The paper presents transport system design guidelines based on experiences gained with parallel implementations of transport and network layer protocols on transputer networks, an implementation architecture for PATROCLOS based on transputer networks and results of a performance evaluation, which indicate promising throughput values.< >
distributed Arithmetic (DA) is used as a method for efficient implementation of inner product computation, where the coefficients of one vector are fixed. In this paper, we compare different structures for the impleme...
详细信息
ISBN:
(纸本)0780312813
distributed Arithmetic (DA) is used as a method for efficient implementation of inner product computation, where the coefficients of one vector are fixed. In this paper, we compare different structures for the implementation of DA. The area-time tradeoff study includes processors based on 1) vectors with N = 4, 8, 16 or 32 variables, 2) four different adder circuits with and without pipelining and 3) two memory saving techniques. The architectures are implemented in a double metal 1.2 μm CMOS technology within a standard cell environment and are verified by simulations. So, this allows a comparison by means of real values for chip area and computation time.
This paper introduces a partial order connection (POC) protocol. Motivated in particular by multimedia applications, POC is an end-to-end connection that provides a partial order service, that is, a service that requi...
详细信息
The authors present a block data decomposition algorithm for two-dimensional grid problems. Their method includes local balancing to accommodate heterogeneous processors, and they characterize the conditions that must...
详细信息
The authors present a block data decomposition algorithm for two-dimensional grid problems. Their method includes local balancing to accommodate heterogeneous processors, and they characterize the conditions that must be met for their partitioning strategy to be of value. While they concentrate on the workstation network model of parallelprocessing because of its high communication costs and inherent heterogeneity, their method is applicable to other parallel architectures.< >
The author has implemented a set of computational physics codes on a network of IBM RS/6000 workstations used as a distributedparallel computer. He compares the performance of the codes on this network, using both st...
详细信息
The author has implemented a set of computational physics codes on a network of IBM RS/6000 workstations used as a distributedparallel computer. He compares the performance of the codes on this network, using both standard Ethernet connections and a fast prototype switch, and also on the nCUBE/2, a MIMD parallel computer. The algorithms used range from simple, local, and regular to complex, non-local, and irregular. He describes his experiences with the hardware, software andparallel languages used, and discusses ideas for making distributedparallel computing on workstation networks more easily usable for computational physicists.< >
Multiprocessor systems should exist in the larger context of distributed systems, allowing multiprocessor resources to be shared by those that need them. Unfortunately, typical multiprocessor resource management techn...
详细信息
Multiprocessor systems should exist in the larger context of distributed systems, allowing multiprocessor resources to be shared by those that need them. Unfortunately, typical multiprocessor resource management techniques do not scale to large networks. The Prospero Resource Manager (PRM) is a scalable resource allocation system that supports the allocation of processing resources in large networks and multiprocessor systems. To manage resources in such distributedparallel systems, PRM employs three types of managers: system managers, job managers, and node managers. There exist multiple independent instances of each type of manager, reducing bottlenecks. The complexity of each manager is further reduced because each is designed to utilize information at an appropriate level of abstraction.< >
暂无评论