Dear Indrani,
Sure. Hope this will work out well for you.
Yours, Roland
Dear Roland,
Thank you very much for your detailed response. I am chiefly interested in running binary black hole simulations. It turns out that I can only purchase a workstation with my project money. However, as you suggest I will try and get an account in a cluster eventually. In case I face any difficulty at that time I will get back to you.
Thanks and best regards, Indrani
Indrani Banerjee Assistant Professor Department of Physics & Astronomy, National Institute of Technology, Rourkela Odisha-769008, India email: banerjeein@nitrkl.ac.in, indrani.physics1@gmail.com
On Thu, Oct 6, 2022 at 8:06 PM Roland Haas rhaas@illinois.edu wrote:
Dear Indrani,
Sorry for the long delay. We had discussed you question already in last week's Einstein Toolkit call.
Unfortunately there is no direct answer that one can provide.
The size of the workstation is very much dependent on the intended use, large differences existing for example between purely vacuum and simulations including matter. The latter usually being at least twice as expensive compute time and memory wise.
Generically a single workstation is not really enough for production type work, since you usually need to either run many smaller simulations (so a single workstation is not enough) or one large simulation (in which case it won't fit on the workstation).
Usually a workstation would be used for development and testing. For development you would require only a "typical" workstation and 16GB of RAM would be enough. For testing, one with large memory (>64GB) can be useful to test code (fits memory-wise0 but will be slow due to only a single CPU socket. I use my own workstation (96GB, 12 slow cores), for this.
You can take a look at the gallery examples:
BBH: https://urldefense.com/v3/__http://einsteintoolkit.org/gallery/bbh/index.htm... 98GB, somewhat reasonable resolution but "real" ones could well be about 2 times more expensive memory wise
BNS: https://urldefense.com/v3/__http://einsteintoolkit.org/gallery/bns/index.htm... 8.8GB for the BNS example but it is *way* too low resolution. Normally the resolution across the star should be about 150m (about 0.1Msun) but currently is 416m (0.28) so 2.8 times too coarse. This will make the "real" run use ~20 times more memory (so ~128GB or so).
For actual data production accessing a small cluster is usually best. Either on campus or via a national or international institution. I do not know what the situation in your location is like. In the US / EU one would usually try ACCESS / PRACE startup allocations first.
Yours, Roland
Hello everyone,
I am planning to buy a workstation from my project. Can anyone suggest
what
should be its minimal configuration so that I can install Einstein
Toolkit
in it and use it to generate numerical waveforms?
Thanks in advance. With best regards, Indrani
Indrani Banerjee Assistant Professor Department of Physics & Astronomy, National Institute of Technology, Rourkela Odisha-769008, India email: banerjeein@nitrkl.ac.in, indrani.physics1@gmail.com
-- My email is as private as my paper mail. I therefore support encrypting and signing email messages. Get my PGP key from https://urldefense.com/v3/__http://keys.gnupg.net__;!!DZ3fjg!6k79Jua84TA1VHo... .