Parallel database systems are increasingly being deployed to support the performance demands of end-users. While declustering data across multiple nodes facilitates parallelism, existing data placement may no longer be optimal due to skewed workloads and changing access patterns. To prevent performance degradation, the placement of data must be reorganized, and this must be done on-line to minimize disruption to the system. In this paper, we consider a dynamic self-tuning approach to reorganization in a shared nothing system. We introduce a new index-based method that faciliates fast and e#- cient migration of data. Our solution incorporates a globally height-balanced structure and load tracking at di#erent levels of granularity. We conducted an extensive performance study, and implemented the methods on the Fujitsu AP3000 machine. Both the simulation and empirical results demonstrate that our proposed method is indeed scalable and e#ective in correcting any deterioration in system throughput. 1.

Parallel discrete-event simulation research has focused mainly on designing efficient parallel simulation protocols. However, the exploitation of parallel simulation technology in realdire applications has been hindered mainly by the lack of simulation support tools. This paper describes the design of $?aDE$ (Structured Parallel Discrete-Event Simulation), a parallel simu- lation environment for developing portable simulation models, and a platform for design experimentation of parallel simulation synchronization protocols. An implementation of the environment, SPaDES/C++, cleanly separates simulation modeling and programming from the details of parallelization such as parallel simulation synchronization and parallel programming. For ease of portability and modular design, SPaDES/C++ is implemented as a parallel simulation library. A comparison of SPaDES/C++ with CSim and Simscrip; using two application examples is discussed.

A new formal approach based on partial order set (poset) theory is proposed to analyze the space requirement of discrete-event parallel simulation. We divide the memory required by a simulation problem into memory to model the states of the real-world system, memory to maintain a list of future event occurrences, and memory required to implement the event synchronization protocol.

ce of my life. iv Preface The emergence of Java programming language together with the Internet creates a new exciting area in the field of simulation, web-based simulation, allows us to develop dynamic simulation models and makes simulation models widely available over the Internet. The main objective of this thesis is to develop a simulation library in Java to make construction, distribution and execution of simulation models easy. The thesis states the motivations and objectives of this project, introduces basic simulation concepts, provides an overview of the major features of the Java programming language and its importance in web-based simulation and presents JSIM, a Java-based simulation and animation foundation library. A bank simulation example developed in JSIM is also discussed. The JSIM library includes six packages: queue, statistic, variate, event,<F

Developing a parallel discrete-event simulation from scratch requires an indepth knowledge of the mapping process from the physical model to the simulation model, and a substantial effort in coping with numerous parallelism issues in the underlying synchronization protocols adopted. The lack of software tools and environments to reduce the development effort significantly is a major hindrance in adopting parallel simulation technology. This paper presents an overview of the SPADES (Structured Parallel Discrete-Event Simulation) scalable parallel simulation framework. We focus on the performance analysis of SPaDES/C**, an implementation of SPaDES on a distributed-memory Fujitsu AP3000 parallel computer. SPaDES/C.. hides the underlying complex parallel simulation synchronization and parallel programming details from the simulationist. We study various ways of improving SPADES execution performance including periodic checkpointing of simulation states, aggregation of messages for logical processes that reside on the same physical processors, and increasing the computational granularity of run-time processes to reduce the costs of synchronization and communication. Our empirical results show that the SPaDES framework can deliver good speedup for applications with large problem size and is scalable.

This paper presents an alternative approach to the parallel implementation of simulation models, based on employing Field Programmable Gate Arrays (FPGAs) as the principal computational elements. FPGAs are digital integrated circuits that combine the benefits of user programmability, fast operation and high levels of circuit integration. It will be shown that this new approach provides significant speed-up over software simulation while avoiding the problems of software implementations of parallel discrete event simulators. The use of FPGAs can form the basis of a highly versatile and interactive simulation environment for general queue simulation. The approach lends itself to the efficient design, maintenance and extension of queue networks. Furthermore as technology improves, the relative performance advantage of FPGA simulators will continue to improve also

Morning commuters adjust their departure times in response to day-to-day changes in congestion. Advanced Traveler Information Systems (ATIS) may enable motorists to employ fundamentally new strategies when adapting their departure times to fluctuations in congestion. At the same time, new driver strategies will likely give rise to different road network behaviors. This paper explores the mutual feedback between driver strategy and traffic system performance through a simulation model of rush hour commuting. Motorists in this model choose departure times according to three adaptive strategies. When commuters apply adaptive strategies that require ATIS in the present model, outcomes for both individual motorists and the system as a whole are by several measures worse than when drivers use a simple strategy that does not require ATIS. These results largely agree with an earlier study of a nearly identical model of rush-hour commuting. This document is available in HTML on the ...

This paper describes a novel method of building multicast trees for Real Time Traffic with Quality of Service constraints. There is a wide range of heuristics to calculate the optimal multicast distribution trees with bounds on the maximum delay from the source to all members. However these heuristics require all the members to be known in advance and assume the existence of a centralized service. We present a heuristic - Best Cost Individual Join (BCIJ) - that joins members one by one, randomly to the existing tree. The method doesn't need previous knowledge of the group members. Trees are dynamically built when each member arrives in the group. A distributed method - Multiple Metric Broadcast (MMB) - for nodes to obtain the best valid path to the existing tree is also presented. MMB is inspired by Reverse Path Forwarding and broadcasts queries to the network that reach existing on-tree members. Theses reply with the best valid paths to the joining member. The member then selects the best path. This avoids the use of any centralized service and the need for link-state information to be available in any node. Evaluation presented shows that the BCIJ produces trees with better cost than existing centralized heuristics and that MMB doesn't have a major effect on the network if the group participation is sufficiently large. 1