High performance clusters (HPCs) based on commodity hardware are becoming more and more popular in the parallel computing community. These new platforms offer a hardware capable of a very low latency and a very high t...
详细信息
High performance clusters (HPCs) based on commodity hardware are becoming more and more popular in the parallel computing community. These new platforms offer a hardware capable of a very low latency and a very high throughput at an unbeatable cost, making them attractive for a large variety of parallel and distributed applications. With adequate communication software, HPCs have the potential to achieve a level of performance similar to massively parallel computers. However, for parallel applications that present a high communication/computation ratio, it is still essential to provide the lowest latency in order to minimize the communication overhead. In this paper, we are investigating message aggregation techniques to improve parallelsimulations of fine-grain ATM communication network models. Even if message aggregation is a well-known solution for improving the communication performance of high latency interconnection networks, the complex interaction between message aggregation and the underlying communication software is often ignored. We show that message aggregation must carefully take into account the characteristics of the communication software to be efficient on an HPC. This methodology can be applied as a preliminary step to tune a message aggregation algorithm for a given combination of hardware architecture and communication software layer.
We present a software approach, namely fast-software-checkpointing (FSC), to reduce the running time of the state saving protocol in optimistic parallel discrete event simulation. The idea behind FSC is to use the ins...
详细信息
We present a software approach, namely fast-software-checkpointing (FSC), to reduce the running time of the state saving protocol in optimistic parallel discrete event simulation. The idea behind FSC is to use the instructions performed during the execution of an event as part of the state saving protocol, hence the total number of instructions due to state saving is reduced. Under FSC the time for saving the state of a logical process prior to the execution of an event E requires an amount of time proportional to the amount of state variables not updated by E's execution, as only these variables must be copied. This outlines that FSC shows some dualism with respect to incremental state saving. We show, however that there exists a basic difference between the two solutions as in FSC some of the state saving instructions are actually event routine instructions, while in incremental state saving they are only added and mixed to the latter ones. We also present a simple software architecture to support FSC and simulation results to demonstrate the effectiveness of such solution. The obtained data show that FSC, combined with a sparse state saving strategy may represent the best checkpointing solution in case of both medium/small state granularity simulations and large state granularity simulations even with small (but non-minimal) portions of the state updated by event execution. FSC may result therefore suited for a wide class of simulation problems.
The Extended Air Defense Testbed (EADTB), is a comprehensive, high- and mixed-level-of-detail, environment for modeling weapon system entities and interactions. Due to the complexity of the models and large scenario s...
详细信息
The Extended Air Defense Testbed (EADTB), is a comprehensive, high- and mixed-level-of-detail, environment for modeling weapon system entities and interactions. Due to the complexity of the models and large scenario sizes, in its current single-threaded form, EADTB is limited in run-time speed. Our goal is to speed up the simulation without re-architecture or re-implementation of the models which comprise 1.76 million lines of Ada code, and without altering model behavior or compromising repeatability and causality. Our work demonstrates that the use of optimistic scheduling techniques and its derivatives, offers the best alternative for object-based systems like EADTB. Specifically we have retrofitted and integrated the same representative pseudo-EADTB prototype with two different object-oriented optimistic scheduling engines (SPEEDES and TEMPO/Thema). We discuss the required architectural and behavioral features of a simulation to allow this retrofit, the issues of C++ to Ada language interfaces, and the employment of the basic services of the optimistic scheduling engines in this environment. Experimental results suggest that order-of-magnitude speed-up is feasible through parallelization, and is scalable to larger experiments simply by adding hardware.
This paper describes a variant of the actor model suited to the development of multimedia systems. The actor model centers on non-overkilling concurrency and customizable constraint-directed scheduling. A distributed ...
详细信息
parallel discrete event simulation algorithms are usually based on time stamp ordering of events. distributed virtual environment (DVE) applications such as DIS typically use unordered event delivery. A partial order ...
ISBN:
(纸本)9780769501550
parallel discrete event simulation algorithms are usually based on time stamp ordering of events. distributed virtual environment (DVE) applications such as DIS typically use unordered event delivery. A partial order called approximate time (AT) is proposed to order events in both domains, facilitating reuse of simulations across DVE and analysis applications. A variation on AT-order called approximate time causal (ATC) order is also described. Synchronization algorithms to realize these orderings are presented as well as performance measurements on a workstation cluster. A long-term goal of this work is to use AT and ATC order to exploit temporal uncertainty in the model to achieve efficient conservative parallelsimulation despite little or no lookahead, a longstanding problem in the field.
This paper addresses the need for reliable multicast (RM) network support for the class of distributed virtual simulations (DVS) that operate with human participants who experience a virtual representation in real tim...
详细信息
The proceedings contain 45 papers. The topics discussed include: security issues in mobile agent technology;ephemeral Java source code;applying computational intelligence for congestion avoidance of high-speed network...
ISBN:
(纸本)076950468X
The proceedings contain 45 papers. The topics discussed include: security issues in mobile agent technology;ephemeral Java source code;applying computational intelligence for congestion avoidance of high-speed networks;routing a class of concatenated omega networks;priority scheduling and buffer management for ATM traffic shaping;from remote objects to physically distributed objects;behavioural specification using XML;using WAP as the enabling technology for CORBA in mobile and wireless environments;Internet - Intranet - Infranet: a modular integrating architecture;dynamic architectures and architectural styles for distributed programs;distributed information systems tailorability: a component approach;integration of object-oriented software components for distributed application software development;design of a group membership service on top of a distributed JAVA actor platform;beyond the whiteboard: synchronous collaboration in shared object spaces;rule-mitigated collaboration technology;some perspectives on network centric distance learning and software engineering;trends in collaboration technologies for net-centric computing;protecting competitive negotiation of mobile agents;incremental security in open, untrusted networks;a simulation model of task cluster scheduling in distributed systems;performance of dynamic load balanced join algorithms in shared disk parallel database systems;and distributed, configurable memory management in an operating system supporting quality of service.
There are at least three major obstacles thwarting wide-spread adoption of parallel discrete-event simulation (a) lack of need, (b) lack of tools, (c) lack of predictability in behavior and performance. The plain trut...
ISBN:
(纸本)9780769501550
There are at least three major obstacles thwarting wide-spread adoption of parallel discrete-event simulation (a) lack of need, (b) lack of tools, (c) lack of predictability in behavior and performance. The plain truth is that most simulation studies can be adequately done on ordinary serial computers. parallelsimulation tools are products of re-search efforts, and simply don't stand up to the demands of modern software engineering. The results of 20 years of research in parallelsimulation reveal it to be a highly complex endevour, with performance results very much dependent on implementation details and model characteristics.
This year's edition of PADS Gestalt therapy provides an opportunity to look around and pick out all the things that are wrong with the PADS community. An account is given on previous assessments and some current d...
详细信息
This year's edition of PADS Gestalt therapy provides an opportunity to look around and pick out all the things that are wrong with the PADS community. An account is given on previous assessments and some current directions in distributedsimulation and their possible relationships to PADS.
暂无评论