In my personal opinion: A Beowulf Cluster (https://en.wikipedia.org/wiki/Beowulf_cluster) built from small machines such as RaspberryPi or BeagleBone would be good for exploring the rudiments of distributed computing or even for a small-scale proof of concept. But, it would not be suitable for carrying out a serious project.

Does your organization have to-be-surplused machines laying around? Can Linux computers be constructed from them? Do you have a spare switch at hand, and some old ethernet cables? It may well be that your IT department has old stuff that nobody wants to use any more. Why not build a cluster from something like that. It would be good for certain serious projects, although not a true supercomputer.

Thanks for your response Peter Raeth

You said that Beowulf clusters are good for small projects but not for serious ones. To illuminate how simple my project is, the most expensive calculation needed a max of 16 GB dynamic memory and took around 15 days. This was neb.x calculation and most of my calculations, which were pw.x, needed 14 GB memory and took 12 hours to complete. Our projects are simple, not like protein modeling. This is why I thought that the 4 x raspberry pi 4 clusters can solve my problem. We prefer experimental studies more than theoretical DFT studies. We only want to support some of our experimental results.

For the second think about IT, our department IT does not have so many spare parts for building new simple machine. They gave most of useful goods to other faculties about a year ago and the remaining parts are older than the universe. Moreover, IT does not even know that the remaining parts are still working. They have some small things like heatsinks, ethernet cables, power suppliers, etc. Thus, I can use them to build clusters.

To sum up, my problems are memory and computer availability. Buying an additional computer will be overkill and expensive. Additionally, I am still designing some other projects which can utilize a raspberry pi cluster.

irkay,

You'd first verfy that your simulator supports the VMD and LINUX version you'll install on Raspberry. Several VMD can be installed on such Raspberry cluster, but may have some problems.

Alternatively, you can upgrade your PC with a better multi-core processor (i7 or i9) and/or more DRAM. You can also use a graphic card , with GPU but be sure first your simulator supports such GPU.

In all cases, take care that you probably installed parallel processing module (MPI), supported by your simulator

Best wishes

Thanks for your response Muhammad Hamza El-Saba ,

To the best of my current knowledge, Raspberry and MPI work best and without problem with QE v.6.5. However, I will check these requirements again to be sure. Moreover, your second option is not suitable for me. I also use my laptop at my home for creating udemy courses, digital arts, and online meetings. I really don't want to increase the burden on my pc with upgrades. I want it to be free from chains of QE.

Berkay-Sungur: I DID NOT say that a Beowulf Cluster would be unsuitable. What I did say is that one built from small computers such as RaspberryPi or BeagleBone would not be suitable.

What you describe would require what I did suggest: A Beowulf Cluster built from reasonable personal computers, a switch, and cables that nobody wants anymore. I've built several of those. The computers do not have to be homogeneous, while I do recommend sticking with all 32-bit or all 64-bit. Whether a computer is a desktop, floortop, or laptop would not matter, as long as they have at least one ethernet port. It is possible to mix machines with different word sizes but that requires additional work to be done.

I've built several such clusters. The first one used machines that were pulled from a scrap pile. Some were put together from parts coming from that same pile. That bunch was all 32-bits of varying CPU speeds. One node was even less than 1mhz. Another was so old it eventually caught fire but was easily replaced. The software would allocate workload to a node according to it's cores processing speed. Node-level software allocated threads according to the number of cores. Thus, the advantages of distributed and shared processing and memory were brought to bear.

Another cluster was built from scrap to mimic the one used in a customer facility that could not be remotely accessed. We rented time on our cluster to a company that needed a remote testbed whose results could be directly installed manually on the customer's facility.

Beowulf clusters can be very handy as a laboratory's computational needs grow beyond what one computer can reasonably handle. All of the clusters I have built required NO capital outlay.