Documentation: https://zhubonan.github.io/disp/
This package is for distributing the workload of ab initio random structure searching (AIRSS) over flexible computing resources. With AIRSS, random yet physically sensible candidate crystal structures are generated and subsequently relaxed. The search is inherently parallel and fast time-to-solution can be achieved by simultaneously running many "small" (a few tens of cores) DFT calculations in a highly efficient way.
In practice, the computing resources can be scattered spatially and temporally - there might be multiple machines that the search can run on, the each job may have to queue and subsequently run for only a limited amount of time.
DISP makes searching easier by providing a centralised server that handles distributing and controlling each search worker, getting real-time analytics data, as well as keeping a database of relaxed structures. It is built upon the original AIRSS package and is fully compatible with the latter.
flowchart TD
A[PC] --> |search tasks| B[Server];
B --> |relaxed geometries| A;
subgraph HPC1
W1;
W2;
end
subgraph HPC
W3;
W4;
end
B --- W1[Worker1];
B --- W2[Worker2];
B --- W3[Worker3];
B --- W4[Worker4];
A newer version of pip ( >21) might be needed - upgrade with:
pip install -U pip
install the package with:
pip install git+https://https://github.com/zhubonan/disp
- A MongoDB server is needed to act as the centralised workflow controller and a place to store search results.
- DISP and its python dependencies needs to be installed on both the local computer and the remote computing cluster/super computer, together with AIRSS.
- Behind the scene, an extended version fireworks is used for workflow management, it should be installed automatically with DISP.
- The plane wave DFT code CASTEP is used for searching, although support for other code can be added if needed.
- Scheduler supported is limited to SGE and SLURM at the moment, but others can be added, given that job information is available at the run time.
Usage is focused on command line, although python interface is also provided for analytics.
Deploy search with:
disp deploy search
Summary of throughput:
disp db throughput
Summary of finished/waiting jobs:
disp db summary
Pull completed search on to the disk:
disp db retrieve-project
- AIRSS code: https://www.mtg.msm.cam.ac.uk/Codes/AIRSS
- Documentation for AIRSS: https://airss-docs.github.io/
- metador - high throughput CASTEP geometry optimisation and analysis: https://github.com/ml-evs/matador
- aiida-fireworks-scheduler - for running standard AiiDA workflow alongside structure prediction workload from this package : https://github.com/zhubonan/aiida-fireworks-scheduler
- Pickard, C. J.; Needs, R. J. High-Pressure Phases of Silane. Phys. Rev. Lett. 2006, 97 (4), 045504. https://doi.org/10.1103/PhysRevLett.97.045504.
- Pickard, C. J.; Needs, R. J. Ab Initio Random Structure Searching. Journal of physics. Condensed matter : an Institute of Physics journal 2011, 23 (5), 053201–053201. https://doi.org/10.1088/0953-8984/23/5/053201.