Seems like that “c” word isn’t going away and I couldn’t be happier. I had the pleasure of seeing what Autodesk has coming in the near future. Two specific projects- Cumulus and Centaur. Both are cloud-powered simulation offerings. Leveraging the client driven interfaces of Moldflow Insight and Inventor respectively.
Project Centaur is a simple structural optimization tool that allows users to setup a very simple Factor of Safety simulation and solver the multiple iterations in “the cloud”. So, the setup is super, super simple.. The interface takes advantage of the slick Inventor 2011 UI, which has one of the more simplified UI for simulation I have ever seen. The slick part is that you simply just hit “optimize” and it packages up the necessary data, solves it, gives you some feedback and then you get a message that the results are ready to go.
Project Cumulus is similar but it focuses on plastic injection parts ala Moldflow. The demo I saw ran a design of experiments simulation that ran a variety of models and presented the optimal solution.
The most amazing thing was that they are using Amazon as the service provider. This is the most exciting point of all. Historically, Amazon had great scaling, but the specs on the individual hardware never really cut it for engineering simulation. But, times are changing. The fact that engineering vendors can access computing power in a traditional commercial environment is awesome! Really awesome.. Total game changer when it comes to cloud computing in the CAE space.
For many engineers that are leveraging simulation to drive design- more is better. Meaning, the ability to run multiple what-if scenarios is critical. This is potentially a new trend or at least the non-typical direction that many take when viewing the “cloud” for simulation. Personally, I think its spot-on.
Historically, many view the cloud with regards to simulation to leverage clusters to solve massively large simulations. This is the typical view in the HPC market.
Neither one of the above Autodesk projects is geared to leveraging parallel computing. Both examples I saw ran numerous simulations in parallel, but each simulation leveraged a single core or at least a single machine.
While, I think this is great and very excited to see design driven simulation in the cloud. The devil is in the details of how this will be packaged and delivered. How will users gain access to the machines, how will they manage the machines in the cloud? Not impossible, but managing and scheduling jobs is non-trivial. Especially if the cloud is elastic where use will fluctuate.
I’m sure the smart people out there have this dialed in, anxious to see the results.
























