Section: New Results

Experimenting with Clouds

Simulation

Participants : Simon Bihel, Martin Quinson.

Providing better interfaces to the users for Cloud Studies. Aware that the current user interface is a impediment to the adoption of our framework by the scientific community, we tried to propose a new, simplified API through the internship of Simon Bihel this summer. We identified several use cases and usage scenario that relevant to our context, and started implementing the new interface that we will provide. This work is still under progress.

Production-ready simulator of large-scale distributed systems. We are currently involved in a complete reorganization of the SimGrid implementation. The goal is two-fold: first we want reduce the tool's learning curve to help beginners. At the same time, we want to normalize the tool's internals so that power users can modify it and/or script the kernel behavior easily. Eventually, we are targeting usages in production and teaching contexts. This long term overhaul is still underway.

Experimentation Testbed

Participants : Anirvan Basu, Julien Lefeuvre, David Margery, Pascal Morillon.

Providing ready to use scripts to deploy popular and complex stacks. The study of complex software stacks on Grid'5000 has always been possible due to the reconfigurability properties of the testbed. Nevertheless, for newcomers with little background in system administration, automating the deployment of these stacks on Grid'5000 has always proved difficult. In 2016, we have provided scripts, that users can fork on github to customise to their needs, to deploy OpenStack, Ceph, Hadoop over Ceph or Sparkle. These have been presented to users during the 2016 winter school.

Use cases

Participants : Deborah Agarwal, Yvon Jégou, Nikos Parlavantzas, Manh Linh Pham, Christine Morin, Kartik Sathyanarayanan, Arnab Sinha.

Experimental Evaluation of Data Stream Processing Frameworks

We worked on evaluating data stream processing environments deployed in clouds. We compared the throughput, latency and energy consumption of Spark Streaming, Storm and Heron real-time data processing environments executed on top of Linux clusters and on top of virtual clusters deployed on top of the OpenStack IaaS cloud. The preliminary evaluation was conducted using the word count application on the twitter data stream. All experiments were conducted on Grid'5000 experimentation platform. The experimental results are described in a technical report to be published in 2017. This work was carried out by Kartik Sathyanarayanan, a student intern in Myriads team in the framework of DALHIS associate team.

Simulation framework for studying between-herd pathogen spread in a region

In our collaboration with Inra in the context of the Mihmes project, we worked on the design of decision tools to evaluate the epidemio-economic effectiveness of disease prevention and control strategies at the scales of the herd, the region and the supply chain. We developed a generic service-based framework to efficiently execute models of infection dynamics in a metapopulation of cattle herds on large-scale computing infrastructures. Our framework has been designed to execute complex regional models combining within-herds epidemiological models. The framework automatically distributes the simulation runs on multiple servers in a cluster and exploits the parallelism of the multicore servers. It relies on OpenMP for parallelizing simulation loops and deals with server heterogeneity and failures. We leveraged PaaSage software stack to deploy the framework on several IaaS clouds.

Mobile application for reliable collection of field data for Fluxnet

Critical to the interpretation of Fluxnet carbon flux data is the ancillary information and measurements taken at the tower sites. The submission and update of this data using excel sheets is difficult and error prone. In partnership with ICOS in the framework of DALHIS associate team, we are innovating the data submission and organization method through a responsive web User Interface able to run on desktop, mobile etc.; thus easing the data lookup and entry process from anywhere including the field sites. Continuing with our initial usability feedback experiences gathered last year on the application interface designs, we decided on the mobile application workflow for implementation. We developed a first prototype based on the PhoneGap (http://phonegap.com/) platform which provided the advantage of the same development code generating mobile application for IOS, Android and Windows platform simultaneously. The main functionality realized in the application prototype is that the user can download all the site data required by logging in through the application; and then view/edit them at the tower site (even in offline mode). The next logical step would be developing the synchronization and validation of data held locally in the application with the servers.