As I’ve already discussed, one of the areas where CAD on the Cloud shows potential is in handling specific tasks that require performing intensive calculations that are suitable for sharing among many processors. That sounds great in theory, and a lot of Cloud marketing (e.g. Virtually Infinite Computing) emphasises that point.

OK, that sounds promising, but how does it pan out in real life? One problem dissuading me from finding out is that Autodesk is being very restrictive with access to many of its Autodesk Cloud products (I’d probably throw a few sample render jobs into the Cloud and compare the performance, but I’m not the right kind of Subscription customer so I’m not allowed). Another problem is that I’m not qualified to review things like structural engineering software where the greatest computational potential appears to lie. Fortunately, Alex Bausk is qualified, so it was interesting to read his review of Autodesk’s Project Storm software.

It’s important to point out here that anything Autodesk with ‘Project’ in the name is not a finished product. It’s an Autodesk Labs thing, designed to attract feedback rather than use in production. I very much approve of this process. It’s one area in which I’m happy to endorse the way Autodesk is approaching the whole Cloud thing, and has several benefits over the flawed private Beta process that Autodesk uses for its mainstream products such as AutoCAD.

The downside for Autodesk when it comes to doing pre-release things publicly is that the criticism can be public, too. For example, selected from Alex’s review:

…the product is, for reasons unknown, available only in selected countries…

…utterly meaningless popups…

Options for analysis settings are, to put it short, appalling.

Project Storm is nothing more than a web envelope for our good old ARSA package. It is basically the same “Robot link” that reviteers have already had for quite a long time…

But the software’s practical use is extremely tiny, to the point of no use at all. You may surely forfeit all hope to do anything with it that would even remotely be relevant to all the “cloud analysis” hype in videos, intros and announcements.

I was unable to make any use of Storm with the sample models that come packed with Revit Structure and Robot Structural Analysis. To feed these default, Autodesk-made models to Storm, some really disruptive editing had to be made that involved deleting whole parts of the model, rendering it practically useless, only able to demonstrate how the process is meant to work.

Ouch! OK, so far it’s mainly just pointing out how half-baked the product is at this stage. Given that it’s a Project and not a finished product, that’s not so bad. It’s shipping products and features that are half-baked that I object to, and Autodesk has certainly produced a few of those. Anyway, here’s the bit I found particularly interesting:

Analysis speed, to a surprise, isn’t looking any good compared to desktop. The Storm’s cloud web analysis is extremely slow, likely because the server would yield a tiny fraction of its resources to your particular task.

In other words, the cloud speed and resource claim in case of Project Storm is no more than a standard cloud computing mantra.

…cloud calculations took around four minutes for this simple model, compared to fraction of a second using desktop…

What does this all mean? It could mean that Alex forgot to turn on the Ludicrous Speed toggle. It could mean that Autodesk is doing this experiment on the cheap and hasn’t paid for enough resources to make it work well. If so, that would be pretty short-sighted, and if Carl wants this Cloud thing to impress people he should sign off on a bunch more cash for Scott’s server farm budget. It could mean that this type of calculation is unsuited to parallel processing, in which case it’s probably not a great candidate for a Cloud product. Or it could mean that the calculation parts of this software haven’t been done properly yet, and everything will fly like the wind as soon as the developers get the hang of things.

Or maybe, just maybe, it means that the reality of Cloud computing isn’t quite as infinitely powerful as the hype makes out.