中国科学院软件研究所机构知识库
Advanced  
ISCAS OpenIR  > 中科院软件所  > 中科院软件所
题名:
并行和分布式仿真的建模和性能分析研究
作者: 胡四泉
答辩日期: 2004
专业: 计算机软件与理论
授予单位: 中国科学院软件研究所
授予地点: 中国科学院软件研究所
学位: 博士
关键词: 并行和分布式仿真 ; 高层模拟体系结构 ; 多分辨率建模 ; 仿真建模 ; 解析性性能模型
其他题名: Research on PADS Modeling and Performance Analysis
摘要: 本文介绍了并行和分布式仿真(PADS)的基础算法,总结了PADS的研究现状,在此基础上,归纳了PADS难以应用推广的原因:对PADS建模方法和工具的支持不够,对PADS性能的影响因素及其关系研究不够深入全面,缺乏方便的性能预测工具。提出了PADS以后研究的方向:结合软件工程的方法,研究现有PADS建模的巨的缺陷,探寻既容纳pADS仿真同步机制又兼容于普通软件设计方法的可操作的实用方法,减少PADS系统设计和实现的额外开销和复杂性;研究包含PADS性能实际影响因素的解析性性能模型和工具,使得用户在并行仿真实现之前能以相对较小的开销较为准确地预测PADS系统的性能。多分辨率建模(MRM)在大型复杂的分布式仿真应用中越来越重要。当前实现的大部分分布式仿真是基于高层模拟体系结构(HLA)来实现的。HLA最初来自于军事领域,现在已被IEEE和OMG接纳为工业标准。本文分析了在采用HLA架构的分布式仿真中的MRM建模的要求。传统的聚集/解聚方法存在时一序一致性问题,因为它在一种分辨率下可能丢弃另一个分辨率情形下的数据。多表征实体(MRE)方法通过将一种分辨率下的属性改变传播到另一种分辨率克服了这种问题,但是存在映射一致性问题。本文提出基于规则的多分辨率建模方法,对不同分辨率的实体属性施加单向映射。该方法操作自然,比MRE方法维护一致性更加简单有效。当前大多数PADS系统的仿真建模通常使用专用的仿真语言或库来描述仿真对象的行为和关系,这样在不同建模框架中建立的仿真模型彼此间缺乏通用性,因而可重用性也较差。统一建模语言(UML)因为具有良好的语法和语义,已成为一种对软件密集型系统的市」品进行可视化、详述、构造和文档化的语言,是当前软件系统设计事实上的标准。然而,标准UML中不包含虚拟时间和时间同步语义,因而缺乏对队Ds系统设计的支持。本文对标准UML语义进行扩展来支持PAOS领域的概念,将扩展后的UML用于PADS系统仿真建模的好处在于利用UML本身的通用性可以降低PADS系统构造的复杂性。主要的工作包括在UM七中引入虚拟时间语义,用包含虚拟时间标记和约束的状态图和交互图来描述仿真实体的功能属性。另外,在PADS系统的设计中,为了保持不同节点的仿真成员间正确的因果关系(cousality),需要指定节点间采用的虚拟时间同步策略。现有的PADS系统开发中,仿真建模不仅要求建模人员熟悉建模对象的领域知识,还要求考虑时间同步算法,使得建模复杂度进一步增加。本文提出“分离法”降低这种复杂度,通过扩展UML语义,创建虚拟时间同步构造型(steeotype),将时间同步细节与仿真模型内部的功能属性逻辑地分离,从而不仅使得仿真对象建模更加容易,而且使仿真模型和时间同步策略两者的重用性大大提高。性能预测对一于并行和分布式仿真的潜在用户来说很重要。好的性能估计模型能够为设计提供指导,减小设计的复杂度。PADS 已有的解析性性能模型往往为了数学推导的方便,丛于一些不实际的假设,有的仅在两个处理器上推导,有的忽略了通讯延迟和状态保存开销。本文用概率方法提出一个乐观仿真的性能模型,基于切合实际的假设,包含了通讯延迟、状态保存开销、非频繁检核、层叠回滚,模型适用于n个处理器。在曙光TC1700并行机上的性能实验表明,我们的解析性性能分析模型得到的结果与实际实现的结果是非常接近的,可以用于乐观仿真实现前的性能预测。
英文摘要: Based on the analysis of parallel and distributed simulation (PADS) basic algorithms and its state-of-the-art, we reason about the difficulties of its widespread application: lack of good PADS modeling methodology and tools; lack of convenient PADS performance prediction tools based on profound understanding of performance elements and they relations. Future directions on the PADS domain are proposed which include research on the existed PADS modeling defects to find out realistic method accommodating the PADS synchronization mechanisms and compatible general software design patterns based on software engineering and research on operatable PADS performance model helping modelers to predict the PADS performance at a accepted cost before implementation. MRM (Multi-Resolution Modeling) is becoming more important as large complex distributed simulations are carried out inside and out of military simulation community. Most current distributed simulations are implemented in the framework of HLA (high level architecture), which was originally defined by US DOD and has been accepted by IEEE and OMG as an industrial simulation standard. This paper analyzes the requirements of MRM in HLA -based simulations. Traditional most adopted practical method "aggregation/disaggregation" has temporal inconsistency problem because it discards the other level as it's at one level. The mainstream theoretic method "MRE (Multiple Representation Entity)" conquers this problem by propagating changes at one resolution to others, but it has weakness in mapping consistence. This paper suggested a doctrine-oriented MRM method. This novel method has good operational easiness and keeps the consistency problem simpler than MRE by enforcing unidirectional map between entities in different resolutions. Currently most parallel and distributed simulation system design employ different modeling methodology and tools in simulation modeling and simulation software system design. However, the entire computer simulation system always is a software system which the de-fact modeling language is UML. This different incurs the increased complexity of PADS construction. We propose UML as a unified language to both simulation object modeling and PADS software system modeling. Our work focus on extending standard UML semantics to support simulation domain specific concepts including introducing virtual time in UML, using UML statechart diagram and interaction diagram with virtual time labels and constraints to describe simulation entity functional property. In addition, we create virtual time synchronization stereotypes to describe PDES time synchronization protocols, which separates the functional properties and synchronization property, thus alleviates the simulation modeling complexity and improves both simulation model reuse and synchronization reuse. Performance prediction is important to PADS potential users. A good performance model can guide design activities and conquer the complexities. Existed analytical performance models are always based on unrealistic assumptions for the purposes of mathematical tractability, such as only two processors are considered or ignorance of communication delays and rollback costs. A probalistic model is proposed for optimistic simulation based on the adherence to realistic assumptions and characterization of various overhead costs, namely non-negligible communication delay, state-saving costs, rollback costs, infrequent checkpointing and rollback costs and n processors. The performance model has been validated against the implementation results obtained on a DAWING TCI700 parallel computer and can be readily used to estimate the performance before the optimistic simulation is implemented.
语种: 中文
内容类型: 学位论文
URI标识: http://ir.iscas.ac.cn/handle/311060/6236
Appears in Collections:中科院软件所

Files in This Item:
File Name/ File Size Content Type Version Access License
LW013932.pdf(2078KB)----限制开放-- 联系获取全文

Recommended Citation:
胡四泉. 并行和分布式仿真的建模和性能分析研究[D]. 中国科学院软件研究所. 中国科学院软件研究所. 2004-01-01.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[胡四泉]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[胡四泉]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

 

Valid XHTML 1.0!
Copyright © 2007-2017  中国科学院软件研究所 - Feedback
Powered by CSpace