Section: New Results

Greening Clouds

Energy Models

Participants : Yvon Jégou, Anne-Cécile Orgerie, Edouard Outin, Jean-Louis Pazat, Martin Quinson.

Simulating the impact of DVFS within SimGrid Simulation is a a popular approach for studying the performance of HPC applications in a variety of scenarios. However, simulators do not typically provide insights on the energy consumption of the simulated platforms. The goal of this ongoing work is to enable energy-aware experimentation within the SimGrid simulation toolkit, by introducing a model of energy consumption for computing applications making use of Dynamic Voltage and Frequency Scaling (DVFS) techniques.

Simulating Energy Consumption of Wired Networks In this work, we aim at simulating the energy consumption of wired networks which receive little attention in the Cloud computing community even though they represent key elements of these distributed architectures. To this end, we are contributing to the well-known open-source simulator ns3 by developing an energy consumption module named ECOFEN. This simulator embeds green levers: low power idle (IEEE 802.3az) and adaptive link rate. An article is currently under review on this topic.

Multicriteria scheduling for large-scale HPC environments Energy consumption is one of the main limiting factor for the design and deployment of large scale numerical infrastructures. The road towards "Sustainable Exascale" is a challenge with a target of 50 Gflops per watt. As platforms become more and more heterogeneous (co-processors, GPUs, low power processors...), an efficient scheduling of applications and services at large scale remains a challenge. In this context, we explore a multicriteria scheduling model and framework for large scale HPC systems. This work is done in collaboration with ROMA and Avalon teams from LIP in Lyon [29], [37].

Dynamic resource management for energy-efficiency The B-Com project, a joint private/public focusing on transfer, targets the design and the implementation of Watcher, a software module used to optimize an OpenStack cloud (in terms of performance, storage optimization or energy savings). This Software module is in the "Big Tent" software development process of OpenStack. In cooperation with Olivier Barais (Diverse Inria Team), we focus on dynamic management of cloud resources for energy-efficiency. Our approach relies on machine learning techniques, models@run-time and dynamic adaptation, and is intended to be included in Watcher. At regular intervals of time, we optimize the use of cloud resources by checking if a better placement of Virtual Machines on physical resources can be achieved, taking into account the migration cost. To achieve this, we have an energy model of the resources which is regularly updated using machine learning techniques that helps optimization algorithms to check if a better configuration can be reached energy-wise. This year we worked on the evaluation of the energy model [28].

Involving users in Energy Saving

Participants : Deborah Agarwal, Ismael Cuadrado Cordero, David Guyon, Christine Morin, Anne-Cécile Orgerie.

Energy-efficient cloud elasticity for data-driven applications Data centers hosting cloud systems consume enormous amounts of energy. Reducing this consumption becomes an urgent challenge with the rapid growth of cloud utilization. An existing solution to lower this consumption is to turn off as many servers as possible, but these solutions do not involve the user as a main lever to save energy. We introduce a system that proposes to the user to run her application with degraded performance in order to promote a better consolidation and thus to turn off more servers. Experimentation results using the Montage workflow show promising outcomes  [47], [48]. We also performed a simulation-based evaluation on how much an energy-aware cloud system could save in energy consumed depending on the proportion of users selecting a green execution mode. These results based on the simulation of two typical daily uses of a data center running 3 real scientific applications will be published in Euromicro PDP 2017.

Energy-efficient and network-aware resource allocation in Cloud infrastructures The ever-growing appetite of new applications for network resources leads to an unprecedented electricity bill, and for these bandwidth-hungry applications, networks can become a significant bottleneck. Towards this end, we proposed microclouds, a fully autonomous energy-efficient subnetwork of clients of the same service, designed to keep the greenest path between its node. This semi-decentralized PaaS architecture for real-time multiple-users applications geographically distributes the computation among the clients of the cloud, moving the computation away from the datacenter to save energy - by shutting down or downgrading non utilized resources such as routers and switches, servers, etc. - and provides lower latencies for users. In this work, we have also analyzed the use of incentives for Mobile Clouds, and proposed a new auction system adapted to the high dynamism and heterogeneity of these systems [20], [19] [46].

Exploiting Renewable Energy in Datacenters

Participants : Sabbir Hasan Rochi, Yunbo Li, Anne-Cécile Orgerie, Jean-Louis Pazat.

Resource allocation in a Cloud partially powered by renewable energy sources We propose here to design a disruptive approach to Cloud resource management which takes advantage of renewable energy availability to perform opportunistic tasks. This Cloud receives a fixed amount of power from the regular electric Grid. This power allows it to run usual tasks. In addition, this Cloud is also connected to renewable energy sources (such as windmills or solar cells) and when these sources produce electricity, the Cloud can use it to run more tasks. The proposed resource management system integrates a prediction model to be able to forecast these extra-power periods of time in order to schedule more work during these periods. This work is done in collaboration with Ascola team from LINA in Nantes  [44], [51][9].

Creating green-energy adaptivity awareness in SaaS application In addition to “green” resource allocation at the IaaS level in Datacenters, we think that users should be involved in “greening” their energy use (SaaS level). We propose that applications should have multiple “modes” of execution, each mode using a different level of energy and providing a different service level. For example, a B2C application may provide more or less recommandations . If this applcation can be dynamically swiched between these modes depending on the availability of green energy, the IaaS can optimize resource allocation better. To enforce his, we have designed green energy aware controllers.

This work is done in collaboration with Ascola team [23], [9].