High-throughput real-time systems require nonstandard and costly hardware and software solutions. Modem, and specially multiprocessor, workstations can represent a credible alternative to develop real-time intensive s...
详细信息
ISBN:
(纸本)1932415262
High-throughput real-time systems require nonstandard and costly hardware and software solutions. Modem, and specially multiprocessor, workstations can represent a credible alternative to develop real-time intensive signal processingapplications. Furthermore, the programming model of Kahn Process Networks (KPN) corresponds completely to this kind of applications and fits perfectly multiprocessor systems. However, the current scheduling of KPN suffers of drawbacks with respect with real-time and efficiency. A new activation strategy of processes in a KPN is presented, which considerably improves the existing techniques. This new activation order takes into account both the usual artificial deadlock detection problem and resolution with time constraints. With this algorithm, it is not required to wait until the entire execution has deadlocked to remove bottleneck. Moreover, an optimized allocation memory and a bounded number of processes context switches mechanism is described.
Peer-to-Peer (P2P) networking uses a proliferate and powerful paradigm that replaces the traditional client/server communication model with an arbitrary communication model where all hosts or peers are both clients an...
详细信息
ISBN:
(纸本)1892512459
Peer-to-Peer (P2P) networking uses a proliferate and powerful paradigm that replaces the traditional client/server communication model with an arbitrary communication model where all hosts or peers are both clients and servers. While P2P is well known for file sharing we developed a P2P system that provides distributed computing. We have developed and currently use a distributed shell over P2P which is a modified version of GNU-Bash that offers the user more computational power by transparently distributing commands over the P2P network. This paper presents the shell over P2P architecture and provides techniques we use to obtain I/O redirection and pipes, parametric computations, cross-file system and heterogeneous computing.
Currently loud data centers exist problems of load imbalance and high power consumption. This paper studies the virtual machine placement policy in cloud environment by applying live migration techniques, and proposes...
详细信息
ISBN:
(纸本)9781538637906
Currently loud data centers exist problems of load imbalance and high power consumption. This paper studies the virtual machine placement policy in cloud environment by applying live migration techniques, and proposes a target host selection algorithm MOGA-THSA based on MOGA. As a heuristic algorithm, through designing excellent genetic operators and fitness functions, it optimizes the load balance and power consumption of the data center with smaller SLA violation rate. The algorithm is implemented on simulation platform CloudSim, and experiments show that it can improve the load balance of cloud data center and decrease total power consumption effectively, thus having a certain guiding significance for researching the virtual machine placement policy.
This paper presents a new Web Transaction processing architecture, namely WebTP based on web services and the X/Open DTP model. Web applications, via the WSDL-style TX interface, delineate global transaction boundarie...
详细信息
ISBN:
(纸本)1932415262
This paper presents a new Web Transaction processing architecture, namely WebTP based on web services and the X/Open DTP model. Web applications, via the WSDL-style TX interface, delineate global transaction boundaries among web sites. WebTP services, via the X/Open-defined interface XA between Resource Managers and the Transaction Manager (Hong Transaction), manager subtransactions in local databases. As a case study of WebTP, a web application called TelePortal is presented to investigate the performance of this architecture. Therefore, WebTP has given a seamless mechanism for embedding the X/Open DTP model in web services to support Web Transaction processing, which provides one promising reference implementation for the future web and Grid transaction services.
In this paper, we will be interested in the study, on a parallel architecture with distributed memories, some techniques of load balancing introduced in the parallelization of Hansen's algorithm [3]: a sequential ...
详细信息
ISBN:
(纸本)1932415602
In this paper, we will be interested in the study, on a parallel architecture with distributed memories, some techniques of load balancing introduced in the parallelization of Hansen's algorithm [3]: a sequential algorithm of Branch-And-Bound (BAB) behaviour, developed for the resolution of the problem of global optimization with inequality constraints. Owing to the fact that this algorithm is dynamic and irregular, we propose parallel algorithms based on the load balancing in terms of quantity and quality of boxes. Our developed techniques are based on the using of "the best first strategy" [2, 7]. Simulations carried out show the importance of the integration of load balancing in order to improve quality of parallel codes. These simulations are based on computing based on the libraries PROFILE/BIAS [4, 5] for the management of interval analysis, C++ and MPI environment for the management of communications between processors of the parallel architecture.
Processor speeds are continuing to climb - but the magnitudes of increase are overridden by both the increasing complexity of software and the escalating expectation of users. Future gains in speed are likely to occur...
详细信息
ISBN:
(纸本)1892512416
Processor speeds are continuing to climb - but the magnitudes of increase are overridden by both the increasing complexity of software and the escalating expectation of users. Future gains in speed are likely to occur through the combination of today's conventional uni-processors to form loosely-coupled multicomputers. In this paper original improvements to the de facto functional language compiler are described. These modifications speed up the execution of functional programs using parallel execution, load distribution, and speculative evaluation. This holistic approach to parallel evaluation has not been published elsewhere.
We are simulating systems with a distributed command function over nodes organized into trusted neighborhoods. An application of this architecture could support a new generation of search engines. Today's Internet...
详细信息
ISBN:
(纸本)1932415262
We are simulating systems with a distributed command function over nodes organized into trusted neighborhoods. An application of this architecture could support a new generation of search engines. Today's Internet search engines use a central command and control function within an Internet environment, whose mark of modernity is the absence of such a centralized structure. We discuss how a query could be answered from distributed data stores using HeteroSort, a sorting technique we described in past papers. A query initiates a cycle of a broadcast to and receives responses from a potentially large number of nodes self organized into an emergent HeteroSort network following the locally determined path of the broadcast. This initial query and the returning responses involve minimal messages. The second cycle consists of the transmission of a more complex query to a specific subset of the initially responding nodes.
Due to the increasing power of PCs, energy consumption has become an important matter and today a subject of study in computing area. The increment of the number of machines in the clusters and the dissemination of co...
详细信息
ISBN:
(纸本)1932415262
Due to the increasing power of PCs, energy consumption has become an important matter and today a subject of study in computing area. The increment of the number of machines in the clusters and the dissemination of computational Grids have become the management of power an essential matter. This paper describes a proposal of Nautilus, a distributed Load Balancing Application for Managing Power in Grids and its strategies to manage power in a computational Grid.
One important piece of system software for clusters is the parallel file system. Two main problems can be found in current parallel file systems and parallel I/O libraries for clusters. One is that they do not use sta...
详细信息
ISBN:
(纸本)1932415262
One important piece of system software for clusters is the parallel file system. Two main problems can be found in current parallel file systems and parallel I/O libraries for clusters. One is that they do not use standard servers, thus it is very difficult to use these systems in heterogeneous environments. Other is that with multiple servers running long time is more probably a fault that blocks clients. There are not a clear solution that integrate, inside the parallel file system itself, the fault tolerance support needed in a flexible way. We propose the use of existing network file servers at the server end of a parallel file system to offer a parallel file system for heterogeneous clusters. By integrating redundant techniques into the parallel file system, is possible to define a fault tolerance support from partition level to file level. This paper describes the design of the fault tolerance support for the next version of the Expand parallel File System. (1) and a preliminar evaluation that has been made in Linux clusters of the first prototype.
Large scale scientific applications such as weather modelling and continuous simulation require the orders of magnitude performance improvement available with the new generation of parallel vector supercomputers such ...
详细信息
Large scale scientific applications such as weather modelling and continuous simulation require the orders of magnitude performance improvement available with the new generation of parallel vector supercomputers such as the Floating Point Systems T Series. Many of these applications exhibit a high degree of parallelism, much of which can be expressed as computational tasks which are of varying size and degrees of dependence on one another, and can be partially ordered for execution. DeSPOT (A distributed Self-Scheduler for Partially Ordered Tasks) is an algorithm for the dynamic distribution of such non-uniform tasks to achieve automatic load balancing on a distributed memory hypercube multiprocessor. This paper describes the DeSPOt algorithm and presents its performance characteristics of various test cases using result timings on the FPS T 20.
暂无评论