Hi Robin,

A fair question - requires a fair answer.

OpenFOAM has been under development for over 15 years now and it is an established piece of software, which removes the need for initial multi-million-dollar development effort. Developing it further is basically funded through consultancy, support and similar projects, as well as training. A substantial part of development and validation has been done by generations of PhD students (Imperial College, Chalmers, Zagreb, Milan, UC Dublin to name the few of early contributors and many new ones) whose model development and validation has been kept alive by the current set-up.

An important development driver is the fact OpenFOAM is reasonably well accepted in academia. It is very nice when the end of a PhD project results not only in a Thesis but also in working code delivered to your industrial sponsor. I personally spend substantial time collaborating with research groups and co-supervising other people's PhD students: the benefit is mutual!

One of the results is what you see: the code evolves to the needs of the end user and (for example) I don't know anyone who was prepared to put down substantial funds for a shiny user interface. Doubtless, this will come, either through a sponsored project or organic development.

Having in mind that market needs is there and the number of "core" developers is rather small and capabilities are pretty hot (see eg. presentations on past OpenFOAM Workshops: http://www.openfoamworkshop.org) it is not difficult to do your sums.

I think there is no need to delve on this further (or jump into self-advertising): ultimately, the CFD market will show us if OpenFOAM is viable or not. The invitation to see it in action in Milan in July is still on

Best,

Hrvoje Jasak

"PS: I priori know that some peoples with closed eyes still say Linux Linux ... but we do not say any becasue fair comparison is not simple"

You're making the rather naive assumption that all computers are pc's and can run windows out of the box. This is definitely not the case - sun, nec, ibm, etc don't even use intel/amd cpus in all their systems; e.g. the Earth simulator uses NEC SX6/8 nodes while roadrunner at los alamos is going to use the Cell broadband engine.

http://www.top500.org/stats/list/30/osfam

Actually the top500 is a rather misleading benchmark - the linpack test is a very atypical operation which scales very well. As I recall the number 1 machine's performance (blue gene) is seriously overestimated for a real world test (i.e. large scale CFD/Meteorology). If a more strenuous test was used, which pushed the memory access bandwidth and interconnect, you'd probably find all the top500 machines owned by banks dropping off the list and machines like the Earth Simulator moving back towards the top.

The Earth Simulator scales really well on "real problems" and I believe, using 70% of the nodes, can sustain around 4Teraflops. However the linpack test is 10 times this! For blue gene the disparity between peak and what is sustainable is supposed to a lot worse (this may have improved in recent years?).

On that note i would like to thank all the participants for taking part in this debate. Thank you very much Dr. Jasak for your fine conclusion which again has established the fact that OpenFOAM is the future. All the commercial code users will be out of work soon. All we need to do is develop pre-processing capabilities for OpenFOAM. I really hate blockMesh!

Soemthing i was able to generate in ICEM in two hours is taking me days to generate using blockMesk, and still its not working. Do you think you can be of any help to me Dr. Jasak?

Thank you all. Have a good day Marco

Folks, A few points I would like to make.

My Background: CFD grunt and support engineer for 13 years at CD-adapco, I used Star V1.8 back in the day. Currently a Fluent and GASP user for a super/hypersonic reseach lab. I'm a user not a developer.

Shiny interfaces: I currently use GASP for all solutions where real gas effects, shock shock interactions, and shock Boundary layer interactions are important. It has, perhaps, the worst user interface I have ever had the displeasure to use. I use that tool in spite of that fact, because it captures the physics of importance. User interface is less important to me than capturing physics. In fact we are looking at a NASA code called VULCAN which has no GUI to speak of.

Physical models: I find myself using Fluent against my better judgement because it has liquid spray models that are necessary for some combustor modeling problems which require that model. I'd like to use GASP but it can't do the job. Physical model selection is important and can drive choices.

Linux: The fastest machines I have access to are some AFRL Cray XT3 machines with 4160 nodes. The compute nodes run "catamount" which I believe is essentially a very stripped down unicos which is linux based. The control nodes run SUSE. The advantage of LINUX is that it can be stripped down to minimize its impact on compute resources. Can you imagine installing windows without getting explorer?

Open Source: 1) I use FLTK, an open source fully free graphics tool kit and find that it is well maintained, documented and supported. Its forums (like this one) are a dynamic resource which generally help me out quickly. Not as fast as GASPs user support, but faster than Fluents.

2) When a bug is found it is reported and then fixed. In the last two years I have not recieved a single bug report from FLUENT. Does anyone seriosly believe there have not been any? I have personally found bugs in GASP. While it was fixed quickly, no report was every passed out to users. Open source users and developers are open about all information, not just the source code. If a problem is found you know it.

Hrvoje: I've had a beer or two with Hrvoje, and so am perhaps less than objective, but I think he is an honorable man, making an honorable living while creating something that, with any luck, may change the way we approach our tool selection.

Anyway I think the "better" debate is a bit silly. Every tool has strengths and weaknesses, and those strengths and weaknesses are different for each user based on the problem at hand, the skill set and specific experience of the user and the constraints of his employer or customer. I hope OpenFoam succeeds simply to keep the comercial guys honest.

My $1.02.

Thanks for listening - Andy R

PS: Those of us who argue better aren't always right. PSS: John C Chien where are you??

Hrvoje.

I have also been developing CFD codes (both academic and commercial) for 15 years.

Out of pure curiosity, I downloaded the code, and have had a look through it as well as the documentation.

My personal opinion is that from a purely architectural standpoint it's a complete and utter disaster. Organizationally it is all over the map, C++ concept usage is over the top, but that is, as I say, my opinion.

However, I have not run the code. If you know what you are doing, it may run great as you say. Whatever. It is not a commercial CFD tool however, so I don't know how you think this.

I really hope you were not working on OpenFoam while you worked for commercial CFD companies. I would think that could get you in a lot of trouble. Talk about conflict of interest.

I guess you would not mind then if I grab a copy, refactor it a bunch so only I can understand it, re-release it under GNU-GPL as OpenFOAM++, and start selling my own consultancy expertise.

Dear Marco,

When I answered the email the first time, I did not realize your inexperience in the area of CFD and that you would not understand a well meant advised.

You just exemplified what companies are trying to avoid: "Something I was able to generate in ICEM in two hours is taking me days to generate using blockMesk, and still its not working. Do you think you can be of any help to me Dr. Jasak? "

You may ask what I mean. Here is the short list:

- Use of an ineffective tool when they can afford one, may be the best one.

- Use of company time to search for support when your boss will always have more important things to do to justify your salary.

- When you find the expert, you must wait for that person to give some of their very valuable time for free just because of their good heart. Hopefully, the company survival will never depend on that project.

- Even if you are willing to pay for support, there is no leverage as a buyer (since you have not bought anything yet) to get the service a paying customer expect.

- You can only get what you pay for!!

Please keep in mind that I have no need to advertise for my company. It already has a department that knows better than I do about the subject and that is what they get paid for.

I always try my best to present an unbiased view in the Main Forum.

Good luck on your studies,

Opaque

Opaque,

you are alright. I confirm (a large portion of this forum members are CFD-novice with little information and idea ). But do you think such comparison is basically correct and goal of opensource is this?

I belive goal of opensource is assistance of researcher to simplify development time, do you think commersial tool could give sufficient development flexibility/service to costomer, or basically do you recoommend development of a pute CFD research (not application) on top of a binary package?

another side is Linux/Windows, you are good candidate to give comment in this regard (i am neigher biased to win nor to Linux):

Could you please let us know what portion of your costomers are Windows user and what Linux/Unix user?

What portion of your costumers use parallel version, what about their cluster type, number of nodes?

What about scalability of commersial codes, do you saw any test on a 1024 cluster?

Thank. John

"- Use of company time to search for support when your boss will always have more important things to do to justify your salary.

- When you find the expert, you must wait for that person to give some of their very valuable time for free just because of their good heart. Hopefully, the company survival will never depend on that project.

- Even if you are willing to pay for support, there is no leverage as a buyer (since you have not bought anything yet) to get the service a paying customer expect."

As a user of one of the big softwares I must add something to your statements. I have seldom received good support although I pay for it. Usually it's a fatiguing process of explaining and then it ends up in a "I'll forward your request to someone that knows" etc. In most cases I hunt for answers on various forums instead of banging my head against some out-sourced support that really doesn't care if I'm a customer or not.

I will say up front I work for one of the commercial CFD codes. I will also say that use of a tool like Openfoam is entirely appropriate for a researcher in a university whose goal is to write new code, perhaps a new turbulence model or a combustion model. What bothers me is that it appears to me that the academic researchers in this forum have no clue or appreciation whatsoever for what CFD analysts in industry do. For the most part they are paid to perform analyses and solve problems as quickly and accurately as possible. Sometimes "quickly" even trumps accurate because an answer good to 1 significant digit today is better than 4 digits in a month when its too late to change a design. Most aren't paid to invent better combustion models or write yet another solver or worry how to port their linux code to windows when the IT department decides one day without asking them to throw away their linux boxes and replace them with windows. When you look at industry, cost of software is not the only thing that matters. The cost of a CFD license is very small portion of all the costs that go into designing a new car or a new chemical plant. People in industry are very sensitive to the entire process that it takes to get an answer. That means how long to clean up dirty cad, how long to mesh it, how long to solve it, how long to post-process it and how long to get a question resolved on the phone if you are having a problem. If they could get this entire process working quickly and reliably in freeware then of course, who wouldn't? They can't. That's why companies like mine have people who know CAD, who know meshing, who know graphics, who understand parallelism, and yes by the way a few people who also know a bit about solving CFD problems and that's the reason people pay for our product.

Yes but...

See my other post and ....

There are times when accuracy does trump expediency even in industry. So onto..

QUESTION 1:

Why do some of us in industry use codes like?:

KIVA,GASP,VULCAN etc etc

GASP has no mesher, a horrible GUI, and a marginal post processer. Thus I use two other commercial products Gridgen and Tecplot to continue my work.

VULCAN is strictly a solver so again a mesher and post processor are required.

Not sure what the current state of KIVA is.

Question 2:

Why do some commercial clients stick with older versions of a commercial code?

From personal experience with a product not to be named, a tourque converter desinger stuck with old version until it no longer ran on his new computers simply because it gave the best match to the physics of interest.

And the answer is:

Sometimes accuracy is the most important thing!

If I have a concern based on personal experience (and I too worked in the CFD industry for a while), CFD almost was still born due to the Colorful Fluid Dynamics syndrome. If it made pretty pictures its gotta be good. Find an old timer who got burned using phoenix back in the beginning.

That said, the commercial tools available do an excellent job on 99% of our problems of interest. But they are expensive. As open source CAD and CAE mature, so will open source FEA and CFD. Look at the current versions of Open Office, Firefox, and Thunderbird. The quality of these codes probably exceeds that of their commercial counter parts. Microsoft Explorer is now copying Mozzila's Firefox!!

Opensource CFD is where Commercial CFD was 30 years ago. But if history is any guide, they will catch up quick. Then things will get very interesting ;-)

- Andy R

Yes but...

See my other post and ....

There are times when accuracy does trump expediency even in industry. So onto..'

>>Sometimes speed is of the essence and sometimes accuracy. I don't accept by the way that commercial codes, properly used, are any less accurate than free codes in general - obviously you can always find a niche code that does something specific better. But you know as well I do that there are ways to cut corners which speed things up at a cost of accuracy. At least the good engineers know what trade-offs they are making. The bad ones don't have a clue.

QUESTION 1:

Why do some of us in industry use codes like?:

KIVA,GASP,VULCAN etc etc

GASP has no mesher, a horrible GUI, and a marginal post processer. Thus I use two other commercial products Gridgen and Tecplot to continue my work.

>>Why don't you go out and try to sell GASP (or even give it away) to a user in an industry with not very much CFD experience and see how far you get? After all its free.

VULCAN is strictly a solver so again a mesher and post processor are required.

Not sure what the current state of KIVA is.

I don't know enough about GASP, but my guess is that it covers a niche area for which those codes provide better answers. Think anyone is ever going to do a moving grid in-cylinder calculation with GASP or HVAC in a big auditorium? And you already say you pay for commercial pre and post-processing. Why don't you find open software for that? I hear Paraview is an open software post-processor.

Question 2:

Why do some commercial clients stick with older versions of a commercial code?

>>There are lots of reasons people stick with old codes (in no particular order).

>>1)They are too lazy to learn something new.

>>2)They have built the old code into their processes, the old code does what they need, and they don't want to re-engineer them for a new code. Our old codes were not generally less or more accurate than the new ones. The equations of CFD haven't changed and the solution methods for simpler physics have been well known for years. Our new codes have more physics available to cover a wider range of problems and they have also been worked over to be easier to use, and more efficient. But

Yes but...

See my other post and ....

There are times when accuracy does trump expediency even in industry. So onto..'

>>Sometimes speed is of the essence and sometimes accuracy. I don't accept by the way that commercial codes, properly used, are any less accurate than free codes in general - obviously you can always find a niche code that does something specific better or maybe contains special physics that are not in a general purpose code. No one code can possibly be best at every last problem. But you know as well I do that there are ways to cut corners which speed things up at a cost of accuracy. At least the good engineers know what trade-offs they are making. The bad ones don't have a clue.

QUESTION 1:

Why do some of us in industry use codes like?:

KIVA,GASP,VULCAN etc etc

GASP has no mesher, a horrible GUI, and a marginal post processer. Thus I use two other commercial products Gridgen and Tecplot to continue my work.

>>I don't know enough about GASP to say, but my guess is that it covers a niche area for which it provides a better answer. Why don't you go out and try to sell GASP (or even give it away) to a user in an industry with not very much CFD experience and see how far you get? After all its free. How long does it take to bring a new user up to speed with it? Think anyone is ever going to do HVAC in a big auditorium or a moving grid in-cylinder calculation with GASP? And you already say you pay for commercial pre and post-processing. Why don't you find open software for that? I hear Paraview is an open software post-processor. If you can't afford solvers, how come there is money for meshing and post-processing? Could it be that even your company can't afford for you to take 2 years to build a mesh with opensource meshing?

VULCAN is strictly a solver so again a mesher and post processor are required.

>>Same question. How long does it take to bring a new user, without any immediate mentor, up to speed. (answer it will never happen).

Not sure what the current state of KIVA is.

Question 2:

Why do some commercial clients stick with older versions of a commercial code?

>>There are lots of reasons people stick with old codes (in no particular order).

>>1)They are too lazy to learn something new.

>>2)They have built and streamlined the old code into their processes, the old code does what they need, and they don't want to re-engineer the process for a new code. Our old codes were not generally less or more accurate than the new ones. The equations of CFD haven't changed and the solution methods for simpler physics have been well known for years. Our new codes have more physics available to cover a wider range of problems and they have also been worked over to be easier to use, more efficient and of course ported to modern platforms. But they are different and I find that some people absolutely refuse to do "different" regardless of how much better it might be.

And the answer is:

Sometimes accuracy is the most important thing!

>>and if you read what I said befre, sometimes speed is the most important - but I didn't say always.

That said, the commercial tools available do an excellent job on 99% of our problems of interest. But they are expensive. As open source CAD and CAE mature, so will open source FEA and CFD.

>>FEA has been around as a CAE tool way longer than CFD. How come there isn't even now a free open source FEA tool that people are clamoring to use? (Probably someone is going to tell me there is).

Look at the current versions of Open Office, Firefox, and Thunderbird. The quality of these codes probably exceeds that of their commercial counter parts. Microsoft Explorer is now copying Mozzila's Firefox!!

>>You say that yet Linux still hasn't made any significant dent in the Microsoft desktop stranglehold and I bet you there are 100 users of Explorer for every Firefox user (and I'm one of the Linux/Firefox/Thunderbird users!). We use some OpenOffice where I work, but its very minor and only because Microsoft hasn't ported Office to Linux.

Opensource CFD is where Commercial CFD was 30 years ago. But if history is any guide, they will catch up quick. Then things will get very interesting ;-)

>>I don't see any significant Opensource use of FEA, or even CAD for that matter - both of which have been around a lot longer than CFD. And if you want to talk expensive - go see what enterprise CAD/PLM tools cost these days. Linux isn't beating Microsoft at very much and that mostly is the niches that Microsoft hasn't aggressively pursued. I'm not worried about my job.

If open source was going to work for engineering simulation tools then there would already be open source CAD for some time. The cad market is at least 10 times larger than CFD, and about 2-3 times as old & established. So, why has open source not taken over the CAD market???

Open source CFD makes perfect sense for academic researchers. It makes no sense for an industrial quality CFD tools. BTW, KIVA , GASP and VULCAN are examples of bad commercial software.

1- http://www.freefem.org/ 2- Google for Salome you will find good products there (French Web Site), sorry I do not have the site address at hand, but I am confident that a good number of French readers will post it. 3- The Cad market is adjusting itself, ProE, Solid Works are feeling the competition from cheap CAD programmes, Alibre www.alibre.com

The open source community is growing, Have no doubt about that, and in all directions, Look for Open Office, Fire Fox, Sea Monkey as examples.

Here is a real story:- One of the best tools for thermal analysis today is the SINDA programme, originally it was developed by NASA scientists, It has no pre or post processing capabilities at all, you had to prepare a text file describing your model following some rules. NASA sold the programme to industrial interests and what is commercially available today (Under the same name) is the same old programme attached to a graphical user interface, and to be accepted, it has to be able to solve the same old text models. Give OpenFoam time, and it will be the main industrial CFD tool, wish this to happen soon.

Dear Ahmed,

Just in case you did not know:

a - SINDA was not developed by NASA, they funded the development. They just rebranded the original version of Chrysler's own CINDA (Chrysler Improved Numerical Differencing Analyzer). I like to give proper credit to those who earned.

b - SINDA is not the same old program as you mention. It has evolved over the years to include new models. C&R is not sitting just billing former users by reselling only.

c - SINDA supports the old files because C&R (current owner) has no choice. SINDA is well embedded into some design systems that will cost a fortune to customers (including NASA, DOD, and their large network suppliers) to upgrade.

Also, SINDA is not alone in the list of pseudo-commercial software out there. There are several old codes (for example, see those initially funded by the U.S. EPA) that have got a new face, and being resold (sometimes with improved performance) by others. The common theme among them is that a goverment made the initial investment (heavy by the way), did a thorough verification process, validated the physical models, and certified the code for certain uses. Therefore, those working in that niche market have no choice but to use the same code. You can buy a different GUI from different suppliers.

I do not disagree with open source. Good stuff has come out of it. However, the fact that is free and that you can modify the source does not make it perfect. I disagree with comments such as "some code" is the best, and the rest is junk. The truth is not black and white. Is it?

Regarding OpenFoam, I thought it has been under development for 10 years max. Nabla Ltd was formed in the late 90's. After 15 years, it still has a long way to go. Good luck for all those along on the ride.

Thanks for the partial support in another response,

Opaque.

Dear Opaque

How can you compare OpenFOAM with a crappy organisation like Nabla Ltd. Even though you run your company but its sad that you have no idea how rapidly OpenFOAM is growing. As i said OpenFOAM has its drawbacks but academic researchers are working on it and very soon something wonderful is gone come out of it. Sooner than you think! We are all contributing to OpenFOAM and there are hundreds of universities working on it around the world.

Opaque,

I partially agree with what you are saying, although I am not dismissive of OpenFOAM as an industrial tool. While it is inevitably not a slick as a commercial package, there are many advantages to it that I can see.

First, the licensing costs of commercial packages prohibit large scale parallel CFD. Using OpenFOAM essentially opens up new possibilities for CFD analysis on an industrial scale (i.e. running 100 jobs in a day or so) that commercial CFD codes simply can't compete with.

Second, I don't think we should hide the fact that most people in the CFD industry feel that they are paying through the nose for what they actually get. Improvements in the code are marginal at best (I didn't notice the difference between CFX10 and CFX11 personally) and technical support is rarely adequate to answer difficult questions. This feeling seems to be growing with the merger of CFX and FLUENT and the expectation of rising license costs (the statement that they are unrealistic goes unchallenged in most conferences I've been to), which is why there seemed to be so much industrial interest at the last OpenFOAM conference.

While I wont be switching to OpenFOAM, I will certainly use it side-by-side other codes with the expectation to switch fully to OpenFOAM within two years or so. I expect I am not the only one thinking this way.

RupertH