Team KerData

Overall Objectives
Scientific Foundations
Application Domains
New Results
Other Grants and Activities

Section: New Results

Distributed random number generator

Participants : Benjamin Girault, Bogdan Nicolae, Luc Bougé, Gabriel Antoniu.

This work has been carried out by Benjamin Girault, student at Ens Cachan , during its Magistère Research Intership, June-July 2009, under the supervision of Bogdan Nicolae, Luc Bougé and Gabriel Antoniu.

The objective was to provide a convenient platform to design a distributed Random Number Generator (RNG). A large number of copies of classical RNGs were run in parallel on the nodes of the Grid'5000 platform. Their outputs are combined into a very large string of numbers using BlobSeer to manage these highly-concurrent accesses to a shared data. This resulted in both high throughput and high unpredictability.

These simulations on Grid'5000 involved up to 415 nodes during a whole week-end, to gather enough data from slow RNGs to analyze them. The volume of data gathered during this week-end was of about 7 MB. For faster RNGs, the program ran until that 16GB of random numbers were gathered, which is the amount needed to efficiently compare RNGs.

Comparing various RNGs showed that the generator which gave the best results in this context is the one based on HAVEGE  [58] , developed at Inria Rennes – Bretagne Atlantique by André Seznec and Nicolas Sendrier.

A description of the experiment together with a statistical evaluation of the RNG quality can be found in the internship report

This work is part of a wider goal which is to create tools to generate, test and evaluate RNGs in a distributed environment, led by Alin Suciu's team at the Faculty of Automation and Computer Science Computer Science, Technical University of Cluj-Napoca (TUCN), Romania.


Logo Inria