I have a few questions regarding the project I am going to start to work on. 1. Its a problem where I have two concentric cylinders the outer cylinder has water flowing through it and the inner cylinder has slurry flowing (mixture of sand and mud and water)

2. Again there is a shaft rotating inside the inner cylinder.

3. I have to see the velocity profile due to the action of all these three.

Can somebody suggest as to what software can be used to model this. If a software can't is it that I need to write my own code. As my professor is establishing the lab recently I need to know how many computers do I need to solve such a problem and the operating system on the computers. Do we need to have a cluster to solve such porblems.

Any help would be appreciated.

Lack of information: Flow rate RPM Diameters What do you need as the output? Velocity profile where?

I think you will find that StarCD will solve this since you can have rotating/moving interfaces. This is not a trivial analysis with any code. With Star you can also add user code if you need to do something at the boundary a little more complex than they supply in the basic code. Yes you will need a cluster if you analysis is anything but a trivial case. LINUX seems to be as good as any O/S but any flavor of UNIX should be able to handle it, maybe also Windows but i've never tried it. 16 processors with 1-2Gb per processor is usually a good number. The new INTEL architectures are pretty quick and you get a lot for your money. This will allow you to solve large cases in 2-3 days. Everything is relative though so you need to determine your workload, model size and any code dependency in order to size anything you will need. If you are looking at a cluster then a head node with the capability of pre/post-processing your models is also nice.

Have a look at the FOAM code also, it is free to academic users and although it supports sliding interfaces it doesnt seem to me as if you will need them unless you have mixing blades on your shaft. The modelling of the slurrey can be accomplished by most codes these days.

Since you are not doing LES, combustion or high Re flows I think your computational requirements will be relatively modest. Depending on your Reynolds number I would think 1 or 2 good dedicated Linux boxes would be enough to cover your needs.