AccaSim: HPC Simulator for Workload Management

current version: 1.0.0

AccaSim is a Workload management [sim]ulator for [H]PC systems, useful for developing dispatchers and conducting controlled experiments in HPC dispatching research. It is scalable and highly customizable, allowing to carry out large experiments across different workload sources, resource settings, and dispatching methods.

AccaSim enables users to design novel advanced dispatchers by exploiting information regarding the current system status, which can be extended for including custom behaviors such as power consumption and failures of the resources. The researchers can use AccaSim for instance to mimic any real system by setting up the synthetic resources suitably, develop advanced power-aware, fault-resilient dispatching methods, and test them over a wide range of workloads by generating them synthetically or using real workload traces from HPC users.

For more information please visit the webpage of AccaSim

What’s new?

  • 12-07-2017 First version of the package.
  • 19-07-2017 Documentation is moved to http://accasim.readthedocs.io
  • 21-08-2017 Automatic plot generation for comparison of multiple workload schedules and benchmark files: Slowdown, Queue sizes, Load ratio, Efficiency, Scalability, Simulation time, Memory Usage.
  • 13-11-2017 Simulating distinct dispatchers under the same system and simulator configuration can be managed under the Experimentation class. It also includes the automatic plot generation.
  • 06-12-2017 A workload generator is available for generating new workloads from existing workloads.

Installation

For standard Python installations, AccaSim can be easily installed using pip:

pip install accasim

The source files are available on AccaSim Git.

Tools

We provide some extra scripts to take advantage of the simulator’s features:

License

This software is part of the results of the research activities of Cristian Galleguillos (cristian.galleguillos.m(at)mail.pucv.cl), who has been supported by Postgraduate Grant (2015 - 2017) by VRIEA-PUCV. The software can be used free of charge and without territorial restrictions under the terms of the MIT License.