Cloud computing now is one of the fastest growing sectors in the IT industry. How will this affect the CAE industry?
Actually, cloud computing is not new for CAE industry, especially for CFD users. 15 or 20 years ago, when you accessed supercomputers in the computing centre from home or office through dial-up modems (do you know what it is?), or ISDN if you are lucky, you were already using “cloud computing”. Now this cloud computing is so popular that every computing event must have a dedicated session for it. If you are not sure, check the coming Supercomputing Carnival 2011 (SC2011) in Seattle in this November or the more geeky one: HPC 2011 in Boston in last April.
Fluent announced the Remote Simulation Facility (RSF) in 2002. This probably is the first commercial CFD cloud computing offer. They upgraded the cluster several times later. The response is, based on community feedback, very good. This e-business model combining software and computing resources is very attractive for small consulting firms, and moderate users. The $3-$20 (depending on the volume of commitment) per CPU hour is quite reasonable, although not so cheap.
It seems ANSYS is quietly retiring this re-branded ANSYS RSolve, although the RSF button is still available in the current Fluent 13. The community forum is live, but not active. In fact, I did not find the signup buttons.
On the contrary, Autodesk recently announced their Autodesk Cloud. But this is mainly for their CAD portfolios, and none of their CAE products (Algor, CFDesign, or Moldflow) is included.
So, what’s wrong with CAE in the cloud?
Some anti-cloud arguments may include:
- Security. Yes, the public cloud is as insecure as your own computing facilities. So, your shared computing resources only give you the false impression of security.
- Reliability. Amazon’s recent downtime in last April indicates the cloud may be evaporated, just like your computer may crash.
- Licensing control. This depends on the licensing agreement whether the license issuer allows you to put the license in the cloud. How to control the access is similar to that in your local cluster.
- Performance. Most cloud computing now relies on virtualization, which affects the performance. This is true, but to what extent? If we talk it 10 years ago, the performance overhead may be 10-30%. but now, it is usually about 3-5% only.
The actual reason for slow or no adoption of cloud computing in the CAE industry is lack of interest from CAE software vendors.
Most CAE vendors rely on license sales and maintenance, not services. Cloud computing is too risky for them because:
- It may hurt the license sales and revenues. For occasional users, it is surely cheaper. This means less money flow into software vendors.
- The profit margin may be not as high as software licensing. You are not Apple, so you need to price your service reasonably unless you want to keep the irons idle.
- It needs considerable investment, financial and technology. It means the executives need to persuade the shareholders and new investors, but most of CEOs and executives are focusing on short-term performance.
- Major accounts (big buyers) are not interested in it, because they afford an idle system of their own. They treat their perpetual license and computing facility as value-adding assets.
CAE in the cloud is not to make CAE fashionable but may push CAE into a bigger market. There are genuine demands from occasional users, from some consulting firms, from small design companies, from those who need occasionally surged capacity.
Unfortunately, due to lack of interest from major CAE vendors, it seems there is a long way towards CAE in the cloud. Now, we have to wait for the shaker, possibly from the second-tier CAE players.
Jeff Waters says
Great article. You will definitely stir the controversy back up as (for some odd reason) people in the CAD/CAE world often treat this like a religious battle.
I contend that most CAE will be cloud based in 10 years. In fact, I think most computing will be.
People who argue against CAE in the cloud generally offer 3 arguments: 1) IP security 2) Vendor lock-in 3) Performance.
I contend that the real issue in IP security is caused by all the unreliable humans. At worst, people are going to sell or steal IP. At best, people are going to be lazy about what they email or communicate to the rest of the world. That is happening today just as it will happen with a cloud based application.
Vendor lock-in is also a straw man. It’s not terribly common to be able to easily port engineering data from one CAD or CAE system to another today. Why should this be a strike against cloud based approaches? Also, if you look at the industries where a trend to the cloud is more established, you will often find that “data liberation” is considered an important selling point. “You own your data, and you can take it with you if you decide to leave.”
As for performance. I think that’s totally valid (today)… and the weakest of all arguments. I haven’t owned a laptop with a built-in telephone modem in many years. Seems like yesterday that every time I got on the internet, I was listening to 10 seconds of screeches, boops, and bops. It wasn’t that many years ago!
Ask yourself this. How much longer do you think there will be XBoxes, Wiis, and the like? If we get just one more generational advance in broadband internet speeds, and all that immersive gaming will be served in real time over the web. Now, THAT is computationally heavy. We’re not far away….
thanks for your valuable input. Yes, CAE or CFD is supposed to embrace cloud computing. But the reality is that it is almost ignored in this industry. What is needed is just a shaker.
Rich Smith says
A CFD industry shaker – now let me see – you mean affordable, unlimited parallel, easy to learn, easy to use, and a cloud-based solver option at no additional cost? Just the strategy we are trying with Caedium.
For our latest news on cloud-based (via Windows Azure) CFD simulation see http://www.symscape.com/news/hybrid-cfd-modeling-cloud-computing
Full disclosure: I’m sure you’ve noticed, but just in case – I represent Symscape the developer of Caedium.
Caedium is a really unique and interesting CFD tool. Cloud computing and GPGPU definitely differs you from all other CFD software vendors.
I hope Caedium can get more and more users…
Victor Solano says
I think this conversation has taken a sharp turn. Check out http://www.autodesk.com/products/sim-360/
Autodesk is delivering real cloud-based, socially-enabled, lightning-fast CAE solutions that are blowing people’s socks off.
Hi Victor, thank you for the update.
Very interesting article. I think it’s great to see that since 2011, when the article was written, a lot has changed. Many new players started offering a cloud-based CAE now.
I think that overall there is definitely a visible movement of CAD and CAE towards cloud, which is opening the door to a better product design to more users. One of the main reasons why we at SimScale (https://www.simscale.com/) decided to build a cloud-based simulation platform were the barriers we saw, that were preventing simulation from becoming a standard tool in product design:
• Access: Both the hardware requirements as well as the licensing policies of traditional on-premise simulation software create a lot of overhead related to installation, maintenance, and update management. This additional IT administration overhead added by dedicated high-performance (HPC) hardware and complex license restrictions of the actual end-user software kept a lot of companies from using simulation.
• Cost: Dedicated hardware, software, as well as training for simulation can generate upfront costs of tens of thousand euros per simulation seat, sometimes even 6-digit investments before the first simulation can be started and its added-value proven.
• Know-how: Due to the large investment required to deploy traditional simulation tools, companies typically made sure to have a team of specialized experts to utilize these tools to their full potential. For this reason, simulation software vendors have continued to develop their tools specifically for experts. These tools are very powerful and complex, requiring physics, domain, and tool specific know-how from the user. As a result, a user who is starting with a traditional tool might need months of training before he or she is able to us the tool effectively.
Our goal at SimScale is to incorporate simulation into the standard engineering software stack by making it more accessible, cost-efficient (you pay only for the simulation capacity you need and can also use a fully free version for publi projects) and for everyone (no need to be a simulation expert).