Boston
May 8-11, 2017

Event Details

Please note: All times listed below are in Central Time Zone

<< Go back
Toward Fog, Edge, and NFV Deployments: Evaluating OpenStack WANwide
Telecom / NFV Operations

Latency is a critical requirement for Internet Of Things and NFV/SDN applications that favors a more distributed Cloud Computing model deployed at the Edge. This model does not cope with the traditional vision that has been driving the development of OpenStack. Although some efforts such as multiple regions and Cells may allow operators to administrate these new infrastructures, there are no studies thatevaluate how OpenStack will behave in such a massively distributed context.


In this talk, we present results of a study focusing on the impact of Wide Area Network communications on real OpenStack deployments. To this aim, we extended the Enos framework to achieve performance experiments while doing traffic shaping. This is a joint work between the Performance and Massively Distributed Working Groups, following the Barcelona presentation. Experiments have been done over the Grid'5000 and Chameleon testbed. Our goal is to gather metrics to provide inputs to other OpenStack WGs.


What can I expect to learn?

While Fog, Edge and NFV infrastructures  are becoming more recognised, the question of using OpenStack to operate such infrastructures, in particular the performance/scalability of the system, is often debated. In this presentation, we will present a reproducible way to evaluate the performance of OpenStack at a WAN level.


Attendees may expect to learn:

  • Current scenarios to operate WANwide infrastructures with the vanilla code (from simple deployments to the multi-region use-case)
  • How to perform reproducible experiment campaigns thanks to the Enos framework?
  • Methodology to emulate network limitations in a WANwide deployment based on traffic shapping
  • Methodology and tools to benchmark an OpenStack deployment.
  • Preliminary results of the impact of the latency/bandwidth criteria on the behaviour of OpenStack
Wednesday, May 10, 2:40pm-3:20pm (6:40pm - 7:20pm UTC)
Hynes Convention Center - Level Three - Ballroom A
View video
Difficulty Level: Intermediate
Tags: Telecom Operator Architect Upstream Scientific Public Cloud
Adrien Lebre
Full Professor
Dr. Adrien Lebre is Professor at IMT Atlantique.  He received his Ph.D. from Grenoble Institute of Technologies in September 2006. His research interests are distributed and Internet computing. Since 2011, he is member of the Architect and Executive boards of Grid’5000. Dr. Adrien Lebre has taken part to several program committees of conferences and workshops ( ICDCS, CCGRID, SC, HPDC,... FULL PROFILE
Ronan-Alexandre Cherrueau
Research Engineer - Discovery Project
Ronan-Alexandre Cherrueau is a Research Engineer at Inria. He received his Ph.D. from Ecole des Mines de Nantes in November 2016. Passionate computer scientist interested in programming, software composition and distributed computing (further information at: https://rcherrueau.github.io) FULL PROFILE
Pierre Riteau
StackHPC Ltd, Principal Engineer
Pierre Riteau is a Principal Engineer at StackHPC, where he participates in the deployment, operation and support of OpenStack clouds for many organisations in the Research Computing field. He is actively involved in the OpenStack community: as the Project Team Lead of Blazar, the official Reservation as a Service component of OpenStack, and as core reviewer of Kayobe, a software project for... FULL PROFILE