Journal of Interconnection Networks

Although data broadcast has been shown to be an efficient method for disseminating data items in a mobile computing system with large number of clients, the issue on how to ensure the data consistency observed by mobile transactions, which are generated by mobile clients, has been largely ignored by researchers in the area. While data items are being broadcast, update transactions may install new values for the data items. If the executions of update and broadcast of data items are interleaved without any control, the mobile transactions may observe inconsistent data values. In this paper, we propose a serialization checking method (SCM) for concurrency control between read-only mobile transactions and update transactions. SCM is based on the framework of an earlier algorithm, Update First with Ordering (UFO), but improves on that algorithm by reducing re-broadcast overhead when the probability of data conflict between updates and data broadcast is high. Simulation experiments have been performed to investigate the performance characteristics of the proposed method.

The monitor concept has been widely used in a concurrent programming environment for implicitly ensuring mutual exclusion and explicitly achieving process synchronization. It has also been extended to support high-level distributed programming. In this paper, we present JDM, a distributed monitor construct in Java for programming large-scale distributed systems. The distributed monitor construct is based on a well-know tree-based distributed mutual exclusion algorithm proposed by K. Raymond. To increase scalability of the construct, a two-level system architecture is developed, where the node level controls the access to the system-wide shared resources using Raymond's algorithm and the process level synchronizes local processes based on the local monitor concept. An object-oriented design of the system architecture is presented. Implementation and results of performance evaluation are reported and factors that influence the performance of the construct are discussed.

We consider routing in hypercube networks with a very large number (i.e., up to a constant fraction) of faulty nodes. Simple and natural conditions are identified under which hypercube networks with a very large number of faulty nodes still remain connected. The conditions can be detected and maintained in a distributed manner based on localized management. Efficient routing algorithms on hypercube networks satisfying these conditions are developed. For a hypercube network that satisfies the conditions and may contain up of 37.5% faulty nodes, our algorithms run in linear time and for any two given non-faulty nodes find a routing path of length bounded by four times the Hamming distance between the two nodes. Moreover, our algorithms are distributed and local-information-based in the sense that each node in the network knows only its neighbors' status and no global information of the network is required by the algorithms.

In this paper, we first study the covergence performance of relaxatio-based algorithms for linear integral differential-algebraic equations (IDAEs), then a parallel decoupling technique to speed up the convergence of the relaxation-based algorithms is derived. This novel technique is suitable for implementation of parallel processing for complicated systems of IDAEs. Factors taking effect on the performance of parallel processing are discussed in detail. Large numerical examples running on a network of IBM RS/6000 SP2 system are given to illustrate how judicious partitionings of matrices can help improve convergence in parallel processing.

The long-range dependence of Internet traffic has been experimentally observed. One issue in handling long-range dependent traffic is how to simulate random traffic data with long-range dependence. The authors discuss a correlation-based simulator with a white noise input for generating long-range dependent traffic data. With the real TCP traffic traces, a simulation model of TCP arrival traffic is empirically developed and the experimental results are satisfactory.

This article addresses the performance of data replication protocol in terms of data availability and communication costs. Specifically, we present a new protocol called Three Dimensional Grid Structure (TDGS) protocol, to manage data replication in distributed system. The protocol provides high availability for read and write operations with limited fault-tolerance at low communication cost. With TDGS protocol, a read operation is limited to two data copies, while a write operation is required with minimal number of copies. In comparison to other protocols. TDGS requires lower communication cost for an operation, while providing higher data availability.

Parallel execution is a very efficient means of processing vast amounts of data in a small amount of time. Creating parallel applications has never been easy, and requires much knowledge of the task and the execution environment used to execute parallel processes. The process of creating parallel applications can be made easier through using a compiler that automatically parallelises a supplied application. Executing the parallel application is also simplified when a well designed execution environment is used. Such an execution environment provides very powerful operations to the programmer transparently. Combining both a parallelising compiler and execution environment and providing a fully automated parallelisation and execution tool is the aim of this research. The advantage of using such a fully automated tool is that the user does not need to provide any additional input to gain the benefits of parallel execution. This report shows the tool and how it transparently supports the programmer creating parallel applications and supports their execution.

Existing user-level network interfaces deliver high bandwidth, low latency performance to applications, but are typically unable to support diverse styles of communication and are unsuitable for use in multiprogrammed environments. Often this is because the network abstraction is presented at too high a level, and support for synchronisation is inflexible.

In this paper we present a new primitive for in-band synchronisation: the Tripwire. Tripwires provide a flexible, efficient and scalable means for synchronisation that is orthogonal to data transfer. We describe the implementation of a non-coherent distributed shared memory network interface, with Tripwires for synchronisation. This interface provides a low-level communications model with gigabit class bandwidth and very low overhead and latency. We show how it supports a variety of communication styles, including remote procedure call, message passing and streaming.

This paper presents a method to cope with fault-tolerant broadcasting in hypercube multicomputers with both node and link faults. The local safety concept is extended to faulty hypercubes with both node and link faults. The local-safety-based algorithm is used in a fully unsafe hypercube, where there is no safe node. A fully unsafe hypercube can be split into a set of maximal safe subcubes. We show that if these maximal safe subcubes meet certain requirements given in the paper, broadcasting can still be carried out successfully and in some case optimal broadcast is still possible. The method is extended to fault-tolerant routing and multicasting when the system contains both node and link faults.

The emergence of mobile computing environments brings out various changes in the requirements and applications involving distributed data and has made the traditional Intelligent Decision Support System (IDSS) architectures based on the client/server model ineffective in mobile computing environments. This paper discusses the deficiencies of the current IDSS architectures based on data warehouse, on-line analysis processing (OLAP), model base (MB) and knowledge based (KB) technologies. By adopting the agent technology, the paper extends the IDSS system architecture to the Mobile Decision Support System (MDSS) architecture. The logical structure and the application architecture of the MDSS and the mechanisms and implementation strategies of the User Access Agent System, a major component of the MDSS, are described in this paper.