On the instruction-level parallel architecture such as VLIW, the performance is affected by the compiler technique. In this paper, we propose an integrated optimization technique which cooperates register reusing, spi...
详细信息
On the instruction-level parallel architecture such as VLIW, the performance is affected by the compiler technique. In this paper, we propose an integrated optimization technique which cooperates register reusing, spilling and rematerialization First, we develop a register allocation method that can be decided, whether the register must be reusing or spilled or rematerialized by the prediction of the execution timing of the instruction in the program, when registers are insufficient. We evaluate our method in comparison with conventional compiler technique for blocks of programs. Second, the spilling and the rematerialization are also applied to the software pipelining to improve the parallelism in the loops. It was shown that the spilling and the rematerialization adopted in the scheduling, improves the parallelism in the loop executions.
distributed storage systems are increasing being used by data-intensive applications for efficient and reliable data delivery. The Network Storage Manager (NSM) is a distributed storage framework with a unique archite...
详细信息
distributed storage systems are increasing being used by data-intensive applications for efficient and reliable data delivery. The Network Storage Manager (NSM) is a distributed storage framework with a unique architecture that maximizes applications control over many of the storage and retrieval policies. Several applications are utilizing NSM for efficient, tunable, and controllable performance. Data layout is one policy that is considered to be application-dependant and tailored algorithms are preferred for application with complex or irregular access patters. Experimental results have shown dramatic performance enhancement when optimized layout policies override the default NSM implementation. Layout algorithms are more effective when proper prefetching and cache replacement policies are implemented.
A wide range of planning applications are combinatorial in nature, making the design of general purpose planning algorithms a still very challenging endeavor. In order to cope with this combinatorial complexity, some ...
详细信息
A wide range of planning applications are combinatorial in nature, making the design of general purpose planning algorithms a still very challenging endeavor. In order to cope with this combinatorial complexity, some of the most recent work in artificial intelligence (AI) planning focuses on the use of sophisticated heuristics, domain search control knowledge, random search and efficient abstract state space encodings such as binary decision diagrams. The additional performance needed by complex planning applications can be provided by adopting massively parallelcomputingsystems, such as networks of clusters. This paper describes a simple, general approach for turning backtrack search based planners into more powerful distributedsystems that run on networks of clusters. Our approach consists in distributing backtrack search points to different processes on the network. We illustrate its potential using DSHOP, a distributed version of the SHOP planner.
Mobile ad hoc networks are difficult to support QoS-driven services because of their unpredictable and frequent topology changes. One of the important concerns is the reliability of routes. Although finding optimal pa...
详细信息
Mobile ad hoc networks are difficult to support QoS-driven services because of their unpredictable and frequent topology changes. One of the important concerns is the reliability of routes. Although finding optimal path for MANET is an NP-complete problem, specific routing protocols sacrifice efficiency for the reliability. In particular, typical reliable routing protocols use redundant paths to achieve reliability. In this paper, we propose a probabilistic model for maximizing the reliability to use for the reliable routing path selection, trying to maximize the connection time.
By adopting code mobility techniques, distributed applications can make better use of network resources. This makes logical mobility particularly attractive in a mobile computing scenario, where the level of network r...
详细信息
By adopting code mobility techniques, distributed applications can make better use of network resources. This makes logical mobility particularly attractive in a mobile computing scenario, where the level of network resources changes continuously. However, it is sometimes difficult to express quantitatively the benefits, and also the overhead, introduced by logical mobility. In this paper, we present the performance evaluation of a migration model where the central component of the application, the server, moves towards the physical barycenter of a network made of mobile hosts. We evaluated both application-level metrics, such as the percentage of delivered messages and communication latency, and network-level metrics, such as the number of routed messages. In many cases, logical mobility contributed to a significant improvement of the considered performance indexes.
A parallel implementation of the bzip2 block-sorting loss-less compression program is described. The performance of the parallel implementation is compared to the sequential bzip2 program running on various shared mem...
详细信息
A parallel implementation of the bzip2 block-sorting loss-less compression program is described. The performance of the parallel implementation is compared to the sequential bzip2 program running on various shared memory parallel architectures. The parallel bzip2 algorithm works by taking the blocks of input data and running them through the Burrows-Wheeler Transform (BWT) simultaneously on multiple processors using pthreads. The output of the algorithm is fully compatible with the sequential version of bzip2 which is in wide use today. The results show that a significant, near-linear speedup is achieved by using the parallel bzip2 program on systems with multiple processors. This will greatly reduce the time it takes to compress large amounts of data while remaining fully compatible with the sequential version of bzip2.
In this paper, we study the auction model for resource management using the SimGrid simulation framework. We investigate three types of auction allocation protocols: (i) First-Price Auction, (ii) Vickrey Auction and (...
详细信息
In this paper, we study the auction model for resource management using the SimGrid simulation framework. We investigate three types of auction allocation protocols: (i) First-Price Auction, (ii) Vickrey Auction and (iii) Double Auction. The goal is to find which one is best suitable for the grid environment from users' perspective as well as from resources' perspective. The results showed that when we consider a mix of risk-averse and risk-neutral users First-Price Auction favors resources while Vickrey Auction favors users. On the other hand the Doubly Auction favors both users and resources.
For interconnection networks, designing shortest path routing algorithms is in general more difficult than designing simple routing algorithms. In this paper, we derive a shortest path routing algorithm for pyramid ne...
详细信息
For interconnection networks, designing shortest path routing algorithms is in general more difficult than designing simple routing algorithms. In this paper, we derive a shortest path routing algorithm for pyramid networks. The proposed algorithm takes O(l) time to determine a shortest path between any two nodes in a pyramid network. We also design a distributed routing algorithm such that an intermediate node takes O(1) time to confirm the next node along the shortest path without any centralized controller.
One implementation of broadcast-based networks is Simultaneous Optical Multiprocessor Exchange Bus (SOME-Bus). It is a low-latency, high-bandwidth, fiber-optic network that directly connects each processing node to al...
详细信息
One implementation of broadcast-based networks is Simultaneous Optical Multiprocessor Exchange Bus (SOME-Bus). It is a low-latency, high-bandwidth, fiber-optic network that directly connects each processing node to all other nodes without contention. To better utilize the communication network and reduce the completion time of a parallel application, this paper describes the Key Message (KM) approach on SOME-Bus clusters. After presenting KM algorithm with SOME-Bus structure, an example is analyzed to evaluate its performance. The analysis result shows improved performance of communication of a parallel application over a system that does not use the KM approach.
As information systems have become ever more complex, the interdependence of these systems has increased;consequently, the issue of survivability has also become increasingly complicated. The need for survivability is...
详细信息
As information systems have become ever more complex, the interdependence of these systems has increased;consequently, the issue of survivability has also become increasingly complicated. The need for survivability is most pressing for mission-critical systems, especially when they are integrated with commercial off-the-shelf (COTS) products or services. In the paper, we identify two survivability models - static and dynamic - and we discuss their trade-offs. The comparison of these trade-offs between the models resulted in the creation of a new hybrid survivability model that provides more robust survivability services than the static or dynamic models can provide by themselves. To prove the feasibility of our ideas, we describe our implementation of each of the three survivability models in the context of a mission-critical banking system in a distributedcomputing environment. Finally, we discuss the lessons we learned from our implementations.
暂无评论