Just to give you an idea in which applications a DEXTER personal computer cluster could be the answer to the need for more calculation power:
- Computational Fluid Dynamics Simulations (CFD),
- Finite Element Analysis (FEA),
- Computet Aided Design (CAD),
- Computer Aided Manufacturing (CAM),
- Computer Generated Image (CGI) processing, rendering,
- Video Editing,
- Code/Software Development,
- and many more.
A living example: the CFD engineer
One of the codes we are using for our simulations has a Pre/Post-processor software tool which runs only on Windows. Using its handy Graphical User Interface (GUI) we can apply boundary conditions, mesh parameters and solver settings. After all is set we run mesh calculations on this Windows node as well.
We use the 3 slave nodes with SuSe Linux to run Solver and save results. We can monitor the whole process either on the Master node or on any computer connected to the local network by accessing Slaves through VNC. When they are done we collect results and post-process them on the Windows node.
A living example: the university lecturer
Imagine the situation where a university has its supercomputer but its computing capacity is almost fully occupied by students. It is so because their thesis projects have the priority. Lecturers would like to run their own simulation projects or develop their own computer codes and want to test them in a multi-node computing environment.
Codes, user defined functions or any complex simulations are famous for not running perfectly right at the first time. When our imaginary lecturer has the window of opportunity to run the newly written code on the supercomputer, well, most probably it will not be perfect. So he has to work more on it and wait again for the next opportunity on the supercomputer to run a new test.
But what if he had his personal cluster at his desk in his office?
His DEXTER personal cluster has four nodes ready to be used as a multi-node test environment to run his code. DEXTER nodes can either work together as a cluster or as four individual workstations if they run different operating systems (OS). This way he can run his code in personal cluster mode (all OS are the same) or on all those different ones without the need for a new installation. He just switches hard drives or use virtual machines.
When he fine-tuned his code, he can use the central supercomputer to run it on 1000 cores or more. It is more likely he will have better results sooner and he made a more efficient use of his supercomputing time this way.
A living example: the CGI artist
It often happens that one software tools is just not good enough for everything. It also often happens that the best software tools for each part of the job need different computer environment meaning that one runs on Windows, the other runs on a different OS.
When a CGI artist has a DEXTER personal cluster, they have four workstations built into a computer case. They can choose to have Windows on two of the workstations with all necessary software tools performing the best on Windows. And they can have an other OS on the other two so that all the best performing programs can utilize the resources of separate graphics optimised workstations. Since all four workstations are connected by a high speed network, files can be quickly moved between them. All four workstations can have their full-size graphics cards, loads of RAM and hard drives. And the whole personal cluster fits beside a regular office desk.