As for the sequential and parallel processing , which exists in the field of cscw , temporal constrains are used to describe the temporal relations between the sequential and parallel tasks . an application , which contains sequential and parallel sub - tasks , is shown as an example at last 同时，文中分析了协同工作中存在的串行和并行工作，采用时态约束描述了串、并行任务间的时序关系，并描述了一个并行协同工作的实例。
The main work of the algorithm extracts the parallelism at all stages , and a dynamic load balancing strategy is used , allocates works to different computation knot in local network . it implements a distributed motional curved solid boolean operation in local network 算法的主要工作是在串行算法的基础上飞提取各个阶段的并行性，并对于不同的并行任务采用不同的负载均衡手段，将任务分布到局域网上不同的处理器节点，以获得实时计算效果。
Firstly , we need consider more in cooperation and message - pass between parallel tasks . secondly , the parallel design pattern is different from that of sing - alone program . thirdly , parallel program has to call complicated apis to accomplish data transfer 首先，并行程序的分析与设计需要更多的考虑各并行任务的协同和消息通信机制；其次，并行程序的编写模式与单机程序不同；最后，并行程序需要调用并行平台的复杂的通信原语来完成数据的传送。
Chapter 4 explores the impact of the task allocation and schedule strategy on the performance of multi - processor system . compares several model methods such as petri net and queue model , then proposes a new model method named distributed finite state machine ( dfsm ) . based on this method , and with the assistant of vhdl design and simulation environment , the model of dynamic reconfigurable multi - dsp system is constructed and the simulation of fft algorithm is run under verilog xl 浙江大学博士学位论文第四章论述了基于模型分析方法的性能研究对并行处理系统设计应用的重要性，结合vvp平台动态可重构并行系统的结构特点以及sharc处理器的特性，对petri网及查询队列等系统建模方法进行比较，提出了分布式有限状态机（ dfsm ， distributedfinitestatemachine ）的建模方法，井对在动态可重构多sharc系统运行的并行任务进行了的建模仿真。
At last , the practical tests indicate that the function characteristics and the performance of this loading system are better than the target we desired . beyond the proper scheduling degree , this loading system of the massive text data based on parallel - job scheduling arithmetic is much better than the classic loading system 最后,本文通过一组测试,从功能和性能的角度说明系统在功能特性和性能指标上,均能达到甚至超过预期目标,在适当的任务并行度的前提下,采取并行任务调度算法的海量文本加载服务的数据加载能力要优于传统的加载服务。
First of all , with the compare and analysis of every solution project , this paper give the architecture of the loading system of massive text data based on parallel - job scheduling , afterward it describes that how to solute the problems such as the high - performance , high availability of loading of massive text data 通过充分比较和分析各种解决方案优劣和适用场合,本文提出了一个基于并行任务调度的面向海量文本数据加载系统的体系结构,并阐述了基于该体系结构设计的加载系统如何解决的诸如海量文本数据加载高性能、高可用的问题等。
On the base of analyzing the characteristics of this massive text data , the thesis using the distributed object middleware , oracle 10g exchange partition feature and parallel - job scheduling arithmetic , and then it designs and implements a loading system of massive text data developed for startpmonitor 本文在分析该类海量文本信息特点的基础上,运用分布式对象中间件、 oracle10g交换分区技术以及并行任务调度算法,从大规模事务处理中间件startpmonitor这个特定应用出发,研究并实现了一种用于面向海量文本数据的加载方案。
