Selecting an Optimal Configuration for Distributed Analysis — Ansoft Designer 7.0 在线帮助文档，Ansys Designer 教程

淘宝官方店推荐课程在线工具联系方式关于我们

	微波射频仿真设计 Ansoft Designer 中文培训教程 \| HFSS视频培训教程套装
	Agilent ADS 视频培训教程 \| CST微波工作室视频教程 \| AWR Microwave Office

Selecting an Optimal Configuration for Distributed Analysis

The combination of Distributed Analysis and Multiprocessing for Distributed Analysis allow for efficient use of your compute resources. Selecting an optimal configuration to maximize throughput is dependent not only on the number of Distributed Analysis and Multiprocessing for Distributed Analysis license bundles you own, it is also dependent on the problem you are attempting to solve.

Each machine you specify in the distributed analysis machine list will use one ansoft_distrib_engine license. If you specify the same machine twice in the distributed analysis machine list, Designer uses two ansoft_distrib_engine licenses on that machine and concurrently solves two parametric variations or frequency points on that machine. This can be a considerable advantage if you have several powerful machines and want to use the full computing power for each machine.

For example, take the situation where you have access to two powerful machines, each with 8 cores on and 64 GB RAM, to be used for distributed analysis. You could specify each machine eight times in the distributed analysis machine list and concurrently solve 16 variations on the two machines, which will use 16 ansoft_distrib_engine licenses. This works well so long as the amount of RAM required to solve each variation or frequency point does not exceed 8 GB RAM (64 GB RAM / 8 cores = 8 GB RAM/core). If the problem requires, say, 16 GB RAM, then it is best to run only 4 variations on a 64 GB machine. In this case, if you send 4 variations to each of your 8-core machine, you will have a total of 8 cores that are not being used while solving.

With the Multiprocessing for Distributed Analysis license bundle, you can use the remaining 8 cores while solving the solution matrix. To do this, set the number of processors to 2 (Simulation Machine Options). Since you have 8 variations running concurrently, you would use 8 ansoft_distrib_engine licenses and 8 ansoft_distrib_engine_mp licenses. If the number of processors is set to any number higher than 1, then the number of ansoft_distrib_engine_mp licenses used while solving is equal to will be equal to the number of ansoft_distrib_engine licenses used. That means that, in the above example, if your problem requires more than 16 GB RAM but less than 32 GB RAM, we would recommend running 2 variations concurrently on each machine and setting the number of processors to 4. From a license perspective, you would use 4 ansoft_distrib_engine licenses and 4 ansoft_distrib_engine_mp licenses.


Copyright © 2006 - 2013 微波EDA网, All Rights Reserved 业务联系：mweda@163.com