Lazy Parsing

When is the right time to break a single model in multiple models?

That is the $25,000 question, isn't it? Do you rely completely on Windows to tell you the size of our Revit database? If so, what size are you afraid of? How is the database compressed?

Most BIMs start as a singe file. Sometimes, you are asked to split out certain parts of our building for others to access (read: structure, medical equipment, etc). But what about the architect's model? You all know by now that there is a connection between the compressed database size (what Windows reports) and the physical memory we have available. You have perfected the 3G switch and maxed out your MoBo with 4gb of the fastest RAM. You have heard and are applying the 10:1 rule (or [10:1]x2 depending on who you ask). You have all increased your virtual memory to be twice the size of our physical. But, with all that in place, what then is the ceiling? What about compresing the file or even re-creating the central file? Sure, your number will drop significantly, but then balloon back up within days. Can you really trust that number?

The obvious answer (which is not an answer at all) is "maybe". Quite honestly, it's the only number we have to go on. There are, however, other things to be considered. How much detail is Revit trying to render onscreen at a time? Is Revit trying to render and\or hide every face of the every family, or have the 3D forms been hidden in plan, leaving the Symbolic lines to represent? How many families are loaded, but unused. Extraneous views, anyone? Embedded jpgs, or, say it ain't so, exploded ACAD! NOOOOOOOOO!!!!! What about other software running at the same time (FOR SHAME!) "File size" is not the only determinate factor in breaking a BIM apart. It is a bit of a red herring. Only through responsible model building and family creation can better performance be achieved. We must all do our part. Tune in later for a look at "Building Responsible Families", where I will try and dissect the family creation process into it's major functional uses, and assess how far is too far enough.

Otherwise, "split the model up when you need to". Unfortunately, there is no magic number.