Job management software for large scale simulations.
A framework for simulation job management: OACIS
Organizing assistant for comprehensive and interactive simulations (OACIS) is an open-source framework for managing your simulation jobs and the results. Using a web interface provided by OACIS, users can handle their simulation jobs in an interactive and intuitive manner. This framework is useful for running large numbers of simulation jobs with a wide variety of parameter sets and recording the results in an organized way. OACIS is free for use under a permissive open-source license.
When conducting research using computational simulations, running large numbers of simulations iteratively to try various models and parameters is common rather than running a single calculation with predetermined models and parameters. When multiple simulations are run using such trial and error techniques, hundreds or thousands of results are commonly obtained. These results are practically impossible to manage manually. OACIS systematically manages simulation jobs and results by automating tedious and error-prone tasks and helps researchers systematically explore parameter spaces.
OACIS provides a simple interface to summarize the simulation results because of which users can quickly inspect simulation results via a web browser. The interface does not comprise complicated file operations or plotting software; therefore, users can quickly create plots to investigate how the simulation results depend on parameters and can concentrate more on essential parts of their research activities. Based on the insights obtained by these inspections, users can repeatedly add jobs with different parameters.
OACIS can run jobs not only on a local PC but also on remote servers. It submits jobs to remote hosts using an SSH connection and automatically checks the completion of the jobs. This is useful for both small-scale computations that can be completed on your own PC and large-scale computations that utilize supercomputers such as the K computer.
All the output files generated by simulation jobs, including the failed ones, are automatically downloaded and retained by OACIS. This feature simplifies debugging. You can quickly start your simulation and then improve it via trial and error.
When using PC clusters or supercomputers, often writing a shell script that conforms to the specifications of the installed job schedulers is necessary. OACIS is designed so that you can customize the script for the job submission using a script language. Therefore, it is configurable for a variety of computational environments.
OACIS provides APIs (Application Programming Interaces) in Ruby and Python to automate operations on OACIS. This is a useful feature for various applications including automatic parameter tuning, optimization, and sensitivity analyses.
