BIM手册中文版（译文）：第二章_2.3.4节-可伸缩性

译文：

很多用户遇到的问题是可伸缩性。当项目模型在实际使用的时候变得非常大时，就会遇到可伸缩性的问题。当运行变得迟钝，哪怕很小规模的运行也会非常费力。建筑模型用掉了电脑很多内存空间。大型建筑会包含上百万个不同形状的构件。可伸缩性同时受建模规模（比如楼层面积）和模型中的详细程度影响。即使在一个简单的建筑中，如果对每个钉子和螺丝都建模，也会出现可伸缩性问题。

参数化建模整合了设计规则，即将某一构件的形状和参数关联到另一个构件的形状和参数中。这些规则形成了关系层次：构件内参数化关系、构件间参数化关系，根据另一个构件的变化调整某一构件的形状，控制轴网和控制曲面之间的层次关系决定了形状的参数和一组相关联构件的放置。虽然构件内和构件间的参数化关系是构件的局部更新，但是这种层次规则的扩散可更新整个模型。局部参数化规则扩散只在模型上产生合理要求，然而某些系统架构却限制了管理这种大批量层次规则传播的能力。同样，把项目分成若干部分来开展的同时仍然管理大批量层次关系是很困难的。

问题就在于内存大小。所有构件形状方面的运行都必须占用内存。管理参数化更新的一个简单方法就是把项目存储在内存中。如此便对可伸缩性产生了挑战，并且对可进行有效编辑的项目模块之规模做出了实际限制。然而，如果规则可在文件之间进行传播，即在某一文件中更新某一构件能致使其它文件自动更新，那么对项目规模大小的限制便会消失。仅有少量专门为建筑开发的应用程序有管理参数化变更在多文件之间传播的手段。我们把必须将所有更新的构件同时存储在内存中的系统称为“基于内存”。当模型大到难以存储在内存中时，就会出现虚拟内存交换，这会造成大量等待时间。其它系统有在文件之间传播关系和更新的手段，并且能在编辑操作期间打开、更新以及关闭多个文件。我们称这些系统为“基于文件”型系统。基于文件型系统运行小型项目一般会有些慢，但是随着项目文件变大其运行速度降低的也慢。

用户将项目分成若干模块已经是一种久经考验的共享工作之手段并且限制了自动更新的规模。参照文件往往也被用于限制可编辑的模块。如果项目中的层次关系不会造成全局项目变更，这些操作是相当有效的。有些BIM工具加强了这些限制。

内存和处理方面的问题，将会随着计算机运行越来越快而自然减少。64位处理器和操作系统也提供了重要的帮助。但与此同时，也会有更加细化的建筑模型和更大规模参数化组方面的需求。可伸缩性方面的问题还会继续存在一段时间。

原文：

A problem that many users encounter is scalability. Problems in scaling are encountered when a project model gets too large for practical use. Operations become sluggish, so that even simple operations are laborious. Building models take a lot of computer memory space. Large buildings can contain millions of objects, each with a different shape. Scalability is affected by both the size  of the building, say in floor area, and also by the level of detail in the model. Even a simple building can encounter scalability problems if every nail and screw is modeled.

Parametric modeling incorporates design rules that relate geometry or other parameters of one object with those of other objects. These come in a hierarchy of relations: within object parametric relations, peer object relations, adjusting one object’s shape in response to the change of another object, and hierarchical relations between control grids and surfaces that determine the parameters of shape and placement of a set of associated objects. While within object and peer object relations update locally, hierarchical rule propagation may generate updates to the whole building. Local parametric rule propagation makes only reasonable demands on models, while some system architec- tures limit the ability to manage propagation of large sets of hierarchical rules.Also, it is hard to partition a project into parts for separate development and still manage a large set of hierarchical rules.

The issue is memory size; all operations on object shapes must take place in memory. The simple solution to manage parametric updates is to carry the project in memory.  This challenges scalability and places practical limits on  the size of a project module that can be effectively edited. However, if rules can be propagated across files, where updating an object in one file can lead to automatic updates propagated to other files, the size limitation of a project disap- pears. Only a few BIM design applications developed especially for architecture have the means for managing parametric change propagation across multiple files. We call systems that must carry all updated objects in memory simultaneously memory-based. When the model gets too large to be held in memory, virtual memory-swapping occurs, which can result in significant waiting time. Other systems have methods of propagating relations and updates across files and can open, update, and then close multiple files during an edit operation. These are called file-based systems. File-based systems are generally a bit slower for small projects but their speed decreases very slowly as project size grows.

User segmentation of projects into modules has been a time-tested way  of sharing work and limiting the scale of automatic updates. Reference files   are often used to also limit what can be edited. These work well if hierarchical relations in a project don’t lead to global project changes. Some BIM tools impose these limitations.

Memory and processing issues will naturally decrease as computers get faster. Sixty-four-bit processors and operating systems also provide significant help. There will be the parallel desire, however, for more detailed building models and larger sets of parametric rules. Issues of scalability will be with us for some time.

版权说明：本文来源BIM译站微信公众号，转载已取得授权

公众号名称：BIM译站

公众号账户：BIM2018130