[Wien] -quota option
Gavin Abo
gsabo at crimson.ua.edu
Sat Jul 1 18:41:27 CEST 2017
Currently, I don't have any experience with that. So I cannot make any
recommendations.
So, you'll likely have to do your own experimentation with it to see
what works best.
It may depend on the Linux file system that you are using. For example,
the maximum file size for ext is 2 GiB [1]. In this case, I think you
would want to start with only a few k-points in each partial vector
files, say XXX = 3, such that the partial vector files contain only 3
k-point. Then, from checking the file size of the partial vector files
with 3 k-points, you could probably estimate how large the partial
vector files would become with more k-points. My expectation would be
that the file size for partial vector files with 6 k-points (XXX = 6)
would be equal to or larger than the combined file size of two of the
partial vector files with 3 k-points. Then, you would want to check
that the partial vector files don't become larger than that 2 GiB
maximum file size or larger than another lower file size limitation
should your system have one.
From the description in the WIEN2k 16.1 usersguide [2], it is my
understanding that -quota essential takes that data that would go in one
large vector file and splits it into several small ones that can be
handled more efficiently by the operating system. You would have to
check if the partial vectors files outputted are stored in a single
folder or distributed on different hard drives of a cluster and are kept
or removed during the calculation. I believe they are kept for post scf
calculations such that the total disk space used for the calculations
doesn't change much, which seems to match with the observation that you
report were it doesn't matter what you used for -quota as it still uses
1 TB of disks space in a single folder even though you should have a
different number of files and sizes for the vector files for the
different XXX cases.
[1] https://en.wikipedia.org/wiki/Comparison_of_file_systems
[2] http://susi.theochem.tuwien.ac.at/reg_user/textbooks/usersguide.pdf
(section "5.6.2 Knight shifts" on page 89)
On 7/1/2017 12:20 AM, Wien2k User wrote:
> Dear Prof P.Blaha and wien2k users;
>
> How to choose XXX k-point for the -quota XXX option because the
> calculation exceeds the capacity of 1 TB
> I tried with XXX = 100 and 999 but the problem persists
More information about the Wien
mailing list