grog 0.91

GeneRic wOrkload Generator

GROG a GeneRic wOrkload Generator

GROG is a generic workload generator (Cloud and HPC like) to create datasets of tasks for scheduling evaluation.

This workload generator is based on a dataset from Google studied in Modeling, classifying and generating large-scale Google-like workload published in the SUSCOM journal (open access version)

The current version only generate traces for the BatSim simulator of DataCenters.

Installation

pip3 install grog

Usage

usage: grog-tool [-h] [--seed SEED] [--offline] [--duration DURATION] [--density DENSITY] [--speed SPEED] output_file

positional arguments:
  output_file          filename to store the workload

options:
  -h, --help           show this help message and exit
  --seed SEED          change seed, default: 42
  --offline            to change from online to offline
  --duration DURATION  duration in days, default: 3
  --density DENSITY    hourly density, default:50
  --speed SPEED        processor density in GHz, default = 1.4

Example

Simple example, uses the default values to generate a file called workload.json

grog-tool workload.json

Longer example, uses some flags to generate a longer duration (here one year).

grog-tool --offline --seed 128 --duration 365 --density 5 long_workload.json

Citation

Please cite (bibtex version) the following if you use this tool

Georges da Costa, Léo Grange, Inès De Courchelle. Modeling, classifying and generating large-scale Google-like workload. Sustainable Computing : Informatics and Systems, Elsevier, 2018, 19, pp.305-314. ff10.1016/j.suscom.2017.12.004ff.

Misc

• This software is under the GPLv3.
• The current point of contact is [e-mail]
• The last version is available here : https://gitlab.irit.fr/sepia-pub/grog
• The current internal API migh evolve, but command line interface should be stable
• The time needed for generating dense and/or long duration traces might be long