A frequently stated advantage of CAD on the Cloud is the access to large amounts of processing power. Instead of relying on your lowly local processor to perform complex tasks, you can instead zap the job up to the Cloud where vast numbers of processors churn away in massively parallel fashion and then zap the results back to you before you’ve even had time to head for the coffee machine.
This is a scenario that applies only for certain types of very complex tasks that are suited to subdividing the calculations among many processors. Autodesk already has a big toe in the waters in several of those areas. The recent Autodesk Cloud changes made available Inventor Optimization, Cloud Rendering, Green Building Studio and Conceptual Energy Analysis to a small subset of Subscription customers. It’s safe to assume that these services will be improved and expanded over the next few years. (Is there anybody out there using Autodesk Cloud services for these processor-intensive tasks? Let’s hear about your real-world experiences.)
What this doesn’t mean is that it makes sense for us all to be using CAD on the Cloud, all the time. The processing time gained by using the Cloud is offset by the communication time spent passing the data back and forth, so any processing gain has to be substantial to make it worthwhile. Twenty years ago, when every zoom extents was followed by a looooong wait, a big swag of extra processing power would have come in very handy. These days, processors are ridiculously fast in comparison. They are also very cheap and getting cheaper. Even low-end PCs have had multiple cores for some years, and these days seeing eight almost unused cores on your performance monitor is pretty normal.
The performance of today’s CPUs and the variable performance of today’s Internet, mean that calculations need to be very substantial to make them worth outsourcing to the Cloud. For the vast majority of tasks associated with using CAD software you simply don’t need to hand the job to somebody else’s hardware, because there is ample capacity right there on your desk.
(As an aside, whether it’s on your desk or a server farm, writing software that takes advantage of all those cores must be really difficult. I say that because today’s CAD software seldom uses more than one or two at the same time. Even a seemingly straightforward split like loading AutoCAD’s Ribbon while allowing you to start drawing appears to be too hard. It took Autodesk four Ribboned AutoCAD releases to even attempt this, and the result is a failure; the cursor lag while background loading the Ribbon is unacceptable.)
For tasks where there is the technical potential to share the load, a remote service still might not be the best solution. How about a private cloud instead, where the processing load gets shared between your company’s idle processors via your LAN, and your data never leaves the premises? It seems to me that such a solution could provide most of the Cloud benefits and remove almost all of the concerns. This has already happened in pilot with some Autodesk software. I’d like to see more emphasis placed on private-cloud-friendly software, because I think it has a much better chance of customer acceptance and the development effort is less likely to be wasted.