Software-based distributed shared memory (DSM) allows multiple processes to access shared data without the need for specialized hardware. However, this flexibility comes at a significant cost due to the need for data synchronization. One approach to mitigate these costs is to relax the consistency model, which can lead to delayed updates to the shared data. This approach typically requires the use of explicit synchronization primitives to regulate access to the shared memory and determine the timing of data synchronization. To circumvent the need for explicit synchronization, an alternative approach is to manage shared memory transparently using the underlying system. While this can simplify programming, it often imposes a fixed granularity for data sharing, which can limit the expansion of the coherence domain and increase the synchronization requirements. To overcome this limitation, we propose an abstraction called the elastic coherence domain, which dynamically adjusts the scope of data synchronization and is supported by the underlying system for transparent management of shared memory. The experimental results show that this approach can improve the efficiency of memory sharing in distributed environments

Shared memory is widely used for inter-process communication. The shared memory abstraction allows computation to be decoupled from communication, which offers benefits, including portability and ease of programming. To enable shared memory access by processes that are on different machines, distributed shared memory (DSM) can be employed. However, DSM systems can suffer from thrashing: while different processes update certain hot data items, the largest amount of effort is spent on data synchronization, and little progress is made by each process. To avoid interference between processes during data updating while providing shared memory at page granularity, more time is reserved for a writer to hold a page in a traditional manner. In this paper, we report on complex thrashing, which can explain why extending the time of holding a page might not be sufficient to control thrashing. To increase the throughput, we propose a thrashing control mechanism that allows each process to update a set of pages during a period of time, where the pages compose a logical area. Because of the isolation of areas, updates on different areas can be performed concurrently. To allow the areas to be fairly well used, each process is assigned with a random priority for thrashing control. The thrashing control mechanism is implemented on a Linux-based DSM system. Performance results show that the execution time of the applications that are apt to cause system thrashing can be significantly reduced by our approach.

随着Web服务使用的广泛,人们普遍发现,Web服务的服务质量(Quality-of-Service,QoS)受网络环境、服务端负载等诸多因素影响不断变化,而保证服务使用过程中的QoS也成为许多Web服务使用者的普遍要求.如何更好地帮助服务使用者选择未来一段时间内符合其服务质量要求的Web服务,同时也帮助服务提供者避免服务质量的违规,是服务计算领域近年来的热点问题.由于ARIMA(Autoregressive Integrated Moving Average Model)模型参数简单并能较好地预测QoS违规,已经在Web服务的QoS预测领域获得了广泛的应用.但是单纯地使用ARIMA模型不能够适应Web服务QoS数据的波动频繁、包含噪声等复杂特点.为了达到更加准确的预测效果,本文提出了一种基于时间序列分析的Web服务QoS预测方法,该方法结合了ARIMA模型与卡尔曼滤波,对服务质量的波动反馈灵敏,较单一的预测模型能够有更准确的预测效果.

Nowadays, computers are worn like wears on wearers, and wearable computers, well known to man, together with a variety of other components, are smart wearable devices. People need computers, but also need close humancomputer interaction. In this demand, wearable computing came into being. Due to the rapid development of computing technology and interdisciplinary technology, the development of wearable computing has been greatly promoted. From the concept and application of wearable computing to start, the development trend was analyzed according to the current emergence of new technological achievements and new signs.

Centralized and distributed evaluation approaches have been proposed for Quality of Services (QoS) measurement. The centralized evaluation approach cannot reflect the user-side QoS and the distributed evaluation approach depend on users to provide evaluation records. In this paper, a hybrid evaluation tool comprising two approaches is proposed. In particular, the centralized evaluation is deployed on a cloud computing platform which is the Amazon web services (AWS). Therefore, the hybrid tool can make evaluation from several AWS regions even if there are no test volunteers. Both the collaborative filtering model and the multiple regression model are implemented in the hybrid evaluation tool for predicting the unknown QoS value. To illustrate the advantages of the hybrid QoS evaluation tool, the scene of a traveler who wants to evaluate and select a best web service in the real world is presented. The results show that the hybrid tool is effective and convenient for users to evaluate the QoS of web services.

Maintenance plays an important role in system reliability enhancement which is an essential requirement for computer systems in service for critical applications. This paper presents a failure number m based maintenance policy for k-out-of-n:G load- sharing computer systems with a goal of maximizing the long-run expected system reward per unit time. Compared to previous study, we take load-sharing into account and employ tampered failure rate (TFR) model to describe failure rate change caused by load- sharing. With some typical numeral examples, we analyzed expected system reward rate under different load stress level and achieved corresponding optimal value of decision variable m. The result shows that component failure has much more impact on heavily loaded system than lightly one, and optimization of maintenance action depends greatly on the load level.

Geographic routing in ad hoc and sensor networks prevails recent years for its efficiency and scalability. The Combined Greedy-Face routing (CGF) is a major family of geographic routing that equipped greedy forwarding scheme with a face routing scheme as a backup scheme to resumes forwarding traffic when greedy forwarding fails. Although the backup scheme enable the routing a deliver guaranteed solution, the inconsistency between two schemes in CGF produce tortuous detour paths that containing excessive nodes. In this paper, we propose a novel geographic routing approach called reactive geographic forwarding (RGF). RGF reactively discovers the relay nodes and constructs paths to destination through the relay node. Simulation results on various network topologies show that RGF significantly improves the performance of geographic routing.

Cloud computing has been widely applied to a wide variety of computing environments, with the traditional distributed computing environment, cloud computing using virtual machines to achieve dynamic resource partitioning. Checkpoint recovery technology is a low-cost method to improve the system availability. This paper analyzes the characteristics of cloud computing, virtualization technology and the checkpoint/recovery technology, a mechanism for integration of virtualization technology and checkpoint/recovery technology has been proposed, this mechanism uses coordinated checkpoint protocols at node level, while uses communication-induced protocols at system level. This scheme is transparent to applications with small performance overhead.

Maintenance policy plays an important role in system reliability and security.With increase on system complexity,Maintenance policies for multi-state system with multiple components come to the focus of research.Maintenance policies are classified into policies for independent components,policies considering components dependency and multi-state model based policies,and introductions are given respectively.

A multi-layered fault injection mode is explored and a multi-layered system robustness testing strategy based on abnormal parameter is put forward. Fault injection methods for three layers (API, DPI and system calls based on workloads) of Linux operation system are designed. And an integrated fault injection platform to multi-layered fault injection for testing robustness of operating system is implemented. Using the multi-layered system robustness testing strategy based on abnormal parameter under the platform, various layers of system robustness evaluation are achieved. Finally, the results under different layers of fault injection are compared and analyzed in order to evaluate the robustness of target system.

This paper proposes a novel multicast routing topology-extended path-and a multicast routing strategy based on it in Large Scale Hypercube networks to reduce the multicast communication of networks traffic and implement parallel transfer. Many optical methods based on multicast path were proposed in past years. However, on the condition that the dimension of hypercube increases, the multicast path produced by these schemes is becoming excessively long and becomes inefficient Extendedpath is a new routing topology constructed by many accessorial paths being appropriately separated from main routing path to shorten the entire routing cost. The novel routing scheme strives to transfer the packets concurrently in main path and extended path to reduce the traffic and make the communication parallel so that the multicast transmission can be sped up in the light of the high regularity of hypercube networks. The simulation results indicate that effectiveness and favorable parallelism of the transmission can be obtained using extended-path strategy. © 2012 Binary Information Press.

At present, multiple rendezvous applications, which use only single-transceiver radios to communicate on multiple channels, are widely used in multi-hop ad hoc networks. However, with increasing demands for communication, it is difficult for nodes to obtain an efficiently wireless channel allocation. The root cause is that some matched transceiver-pairs of nodes must exist that cannot correspond with each other; nevertheless, several channels are available in the network in any previous multiple rendezvous approaches. In this paper, a novel distributed algorithm called MCCE is presented to assign snooping channels for nodes by utilising only the status of neighbours within a 2-hop range of the node. Compared with previous ‘similar fair’ approaches, MCCE can improve the performances of MAC protocols in two conditions. If there is no idle-channel in a node’s 2-hop range, then MCCE will direct a best snooping-channel to decrease the potential of channel contention. If there are mass idle-channels in a node’s 2-hop range, then MCCE will direct a snooping-channel to improve the throughput, similar to non-collision channel assignment approaches, which are NP-complete.

In order to improve the efficiency of the fault injection automation test for high-performance fault tolerant computer.This paper designs a fault injection test automation platform based on the open source automated testing tool Software Testing Automation Framework(STAF).This paper discusses the distributed operation environment and the platform software architecture,and describes the flow diagram of the algorithm.Experimental result shows that the platform can improve the reusability of the test tasks and reduce the testing time.

To overcome the limitation in sample system number and test period during the availability test for a transaction processing-oriented fault tolerant computer, an availability assessment method was proposed and a corresponding assessment system was realized. The availability assessment system consists of a multi-level fault injection platform, an application workloads simulator and an availability assessment toolkit. The fault injection platform is designed for automatically injecting various fault-loads into target systems in batches. The application workloads simulator can generate transactions launched by end-users and send them to target systems as workloads. The availability assessment toolkit is designed for several tests, including reliability relationship test among functional subsystems, reliability relationship test among field replaceable units (FRUs), redundancy test of different kind of FRUs, mean time to recovery (MTTR) test, and availability validation test. The evaluation results of the tests on HP Superdome fault-tolerant server accord with official documents, which proves the effectiveness of the assessment system. This research is important for computer manufacturers to predict availability metric and it is also important for end-users to verify system availability.

针对当前大规模P2P网络失效检测负载对系统可扩展性的影响,对失效检测结果的共享机制展开了研究,提出了一个基于被动订阅机制的低开销失效检测(L-FD)算法.L-FD算法通过被检测节点来建立检测结果的订阅关系,使每个节点只需保持常数个检测关系即可获得所有邻居节点的失效状态,在节点不发生失效情况下可使检测负载下降为O(N).该算法的结果共享关系可灵活建立,不受覆盖网拓扑结构及同步时钟等因素的影响,可灵活适应不同的P2P系统.仿真实验及分析结果证实了这一新算法的可行性和有效性.

By analyzing the indoor office LOS channel model defined by IEEE 802.15.4a standard and deducing the variance for intra-symbol interference (IASI), inter-symbol interference (ISI) and multiuser interference (MUI), performance analysis model for ultra-wideband (UWB) based wireless sensor networks (WSN) is proposed, and the bit error rate (BER) formulation is also obtained. The comparison of performance analysis model with and without intra-symbol interference shows that the intra-symbol interference cannot be neglected and such interference will significantly decrease the system performance. In order to verify the proposed performance analysis model and BER formulation, the semi-definite programming (SDP) algorithm is used to obtain the pulses complying with the spectrum requirement of China and IEEE 802.15.4a standard, and by using such pulses, simulations and theoretical analysis are compared, the comparison results show the validity of the proposed performance analysis model and BER formulation.

Multiple Rendezvous approaches utilized multi-channel which was only with a single transceiver radio may enhance Mac throughputs of multi-channel ad hoc network no less than that of other approaches which need extra radios or hardware for time synchronization. However, for the media-access methods of Multiple Rendezvous, nodes in ad hoc network should be listening on different channels to preserve the control packets of any data transmission with noneinterference each other. Therefore, for any broadcast data packet or multicast data packet, it should be transmitted to its destinations like a set of single-cast data packets. And then other performances of ad hoc network as routing may be reduced by the high overload of broadcast packet. In order to reduce the transmission number of a broadcast packet, a method of using prepare tones (single-cast tone and broadcast tone) before data-packet transmission is presented, we called it PTMC (Prepare Tone for Media-access Control). In the method of PTMC, any broadcast packet will be send out only once to preserve it can reach any idle destination, which is not sending or receiving any data-packets at that time, nevertheless any single-cast packet will be matched and send out just as MAXM done. Simulation results among PTMC, MAXM and SSCH show: If every broadcast packet was send to potential receivers one by one, the throughputs of ad hoc network routed by AODV and media controlled by MAXM or SSCH may not be improved as the available channel number is increasing, for reasons of the routing discovery time increasing. Using PTMC to solve this problem, whatever data-traffics or available channels are provided, the throughput of PTMC will be better than the other two. Thus with available channels increasing, the throughput of PTMA is improved.