With the updates to Fusion 360, which is bringing cloud-based simulation to the mainstream, Al Dean ponders how the monetisation of simulation-based experimentation could affect our design and engineering decisions
Let’s talk about simulation and the cloud; two topics that have been floating around for some time.
While the “cloud” is a relatively recent term, the fact of the matter is that using remote computation capabilities is something that has been linked with simulation and analysis for decades.
Those of us grey haired enough to remember may recall that it was often a data centre or a honking great big server rack that did all the computation, while the lighter (for the times) workstation was where you sat doing the pre and post processing.
What’s changed in recent times is that we’re now being led to believe that this in-house facility is old hat and offloading the computation to a cloud-based server is the future.
For some, that will never be the case. Many large scale OEMs already have data centre facilities in-house and it’s more cost effective to have the solver code run on those existing facilities.
But for smaller organisations, the use of local hardware is an option (and it’s certainly cheaper than it ever has been), but the licensing of simulation products from traditional vendors to run across serious computation hardware is charged by the core or by the CPU hour — and that’s certainly not cheap.
This is where cloud-based simulation technologies have the potential to lower the barrier to entry.
To make the most of simulation’s potential it needs to be used heavily and iteratively without worrying about how much you’re using it. That may not be the case for much longer.
If you want to engage in more complex and iterative forms of simulation — whether it’s non-linear problems, optimisation or anything else — your local workstation probably isn’t the best place to do it.
The simulation will lock it up for hours on end, meaning you can’t get anything else done while the box under your desk chunks away at those numbers.
This is the sweet spot for cloud-computed simulation: You essentially have your own data centre facility but without the overheads.
Your local workstation can keep ticking over nicely, allowing you to carry out all those other tasks you need to do on a daily basis, while the cloud handles all the heavy grunt work, delivering back your results when they’re done.
Not quicker, but certainly more efficiently (in terms of availability of resources) than doing everything locally.
This is the core promise of the simulation capabilities Autodesk is starting to introduce in Fusion 360 (read our Fusion 360 — Q4 2016 review).
Although they are still in Preview mode, these advanced simulation tools are based on the Nastran solver.
Nastran is perhaps one of the most respected and accepted solvers on the market today. Its specialism in both linear and non-linear studies has seen it been adopted across a wide range of industry sectors. The term ‘de facto standard’ is often bandied about, but in this case, it’s true.
But here’s the thing. Autodesk is planning to release these capabilities in a ‘cloud-solve only’ form. That means, if you want to use them, you need to use Autodesk’s server facilities for your computation.
While the basic simulation tasks within Fusion can be solved locally or on the cloud, when it comes to these advanced tools, it’s cloud-only.
As you’ll see from the review, that decision puts a price on each and every time you use the tools.
If you need to just tweak a study because you forget a parameter, there’s a cost attached, but if you spend a few rounds of solves to nail down how your study works best, there’s a cost attached.
Yes, your subscription to Fusion 360 brings you a good amount of computation capability as part of it (1,000 cloud credits, which is enough for a good 40 or so shape optimisation runs), the fact is that whether you’ve shelled out hard cash for it or not, there’s a cost attached.
Blow out that allowance in the first three months of the year and you’ll need to buy more credits and that has a finite, fixed price — 1,000 credits costs $1,000.
Whether or not that’s a good thing is not what piqued my interest — ultimately, the market and user adoption will make that call — what I find fascinating is that we’re now approaching a time where mainstream vendors are looking to monetise our design and engineering processes and decisions, and Autodesk isn’t alone in this.
There are many vendors working on cloud-based simulation, both new start-ups (such as SimScale) or older guard vendors.
While, historically, there was a higher-cost attached to adopting these more complex simulation technologies (in terms of software licensing and hardware), you were free to use them as you saw fit.
Simulation is an incredibly powerful tool, enabling you to investigate a design or engineering project from the very earliest stages in development.
Furthermore, to make the most of simulation’s potential, it needs to be used heavily and iteratively, without worrying about how much you’re using it.
That may not be the case for much longer. Software is often purchased as a capital expense and once bought, you’re free to use it as much as you need.
Whether it’s day in, day out, for 12 hours a day or just once a week, as long as you’re deriving the value you need from it, you’re good.
What doesn’t happen is that every time you load it up and use it, you’re charged by the hour or by the operation.
While this transactional type of arrangement is common in other areas of service provision, we’ve not really seen it in the design and engineering sector as yet.
So I’m wondering how things might change.
Imagine working on a product that’s a clear case for shape optimisation. You set up the study, then realise you’ve blown out your limit on your cloud account.
How easy is it going to be to get the management or purchasing department to sign off another grand’s worth of computation? What delays will your project incur or how much more energy will your new product use unnecessarily because it’s over weight?
Al Dean ponders how monetisation of simulation-based experimentation could affect our designs