Section: New Results

Energy-efficient Resource Infrastructures

Participants : Maria Del Mar Callau Zori, Alexandra Carpen-Amarie, Bogdan Florin Cornea, Ismael Cuadrado Cordero, Djawida Dib, Eugen Feller, Sabbir Hasan Rochi, Yunbo Li, Christine Morin, Anne-Cécile Orgerie, Jean-Louis Pazat, Guillaume Pierre, Lavinia Samoila.

Energy-efficient IaaS clouds

Participants : Alexandra Carpen-Amarie, Christine Morin, Anne-Cécile Orgerie.

Energy consumption has always been a major concern in the design and cost of data centers. The wide adoption of virtualization and cloud computing has added another layer of complexity to enabling an energy-efficient use of computing power in large-scale settings. Among the many aspects that influence the energy consumption of a cloud system, the hardware-component level is one of the most intensively studied. However, higher-level factors such as virtual machine properties, their placement policies or application workloads may play an essential role in defining the power consumption profile of a given cloud system. In this work, we explored the energy consumption patterns of Infrastructure-as-a-Service (IaaS) cloud environments under various synthetic and real application workloads. For each scenario, we investigated the power overhead triggered by different types of virtual machines, the impact of the virtual cluster size on the energy-efficiency of the hosting infrastructure and the tradeoff between performance and energy consumption of MapReduce virtual clusters through typical cloud applications  [45] .

Energy-aware IaaS-PaaS co-design

Participants : Maria Del Mar Callau Zori, Alexandra Carpen-Amarie, Djawida Dib, Anne-Cécile Orgerie, Guillaume Pierre, Lavinia Samoila.

The wide adoption of the cloud computing paradigm plays a crucial role in the ever-increasing demand for energy-efficient data centers. Driven by this requirement, cloud providers resort to a variety of techniques to improve energy usage at each level of the cloud computing stack. However, prior studies mostly consider resource-level energy optimizations in IaaS clouds, overlooking the workload-related information locked at higher levels, such as PaaS clouds. We argue that cross-layer cooperation in clouds is a key to achieving an optimized resource management, both performance and energy-wise. To this end, we claim there is a need for a cooperation API between IaaS and PaaS clouds, enabling each layer to share specific information and to trigger correlated decisions. We identified the drawbacks raised by such co-design objectives and discuss opportunities for energy usage optimizations. A position paper has been published on these aspects [15] . Ongoing work is currently conducted in order to quantify the actual possible gains both energy and performance-wise for this IaaS-PaaS co-design approach.

Energy-efficient and network-aware resource allocation in Cloud infrastructures

Participants : Ismael Cuadrado Cordero, Christine Morin, Anne-Cécile Orgerie.

Cloud computing is increasingly becoming an essential component for Internet service provision, yet at the same time its energy consumption has become a key environmental and economic concern. It becomes urgent to improve the energy efficiency of such infrastructures. Our work aims at designing energy-efficient resource allocation for Cloud infrastructures. Yet, energy is not the only criterion to take into account at risk of losing users. A multi-criteria approach is required in this context to satisfy both users and Cloud providers.

The proposed resource allocation algorithms will take into account not only the computing resources but also the storage and networking resources. Indeed, the ever-growing appetite of new applications for network resources leads to an unprecedented electricity bill for network resources, and for these bandwidth-hungry applications, networks can become an significant bottleneck. This phenomenon is emphasized with the emergence of the big data paradigm. The designed algorithms would thus integrate the data locality dimension to optimize computing resource allocation while taking into account the fluctuating limits of network resources.

In 2014, several experiments were performed to understand and quantify networking energy consumption. These experiments include network protocol energy consumption in the devices, configuration energy consumption in switching/routing devices and associated energy consumption to real cloud computing applications (e.g. Google drive). These experiments have been performed over systems provided by Inria such as Grid'5000 and specific network devices (e.g. level 3 router for a private LAN). Based on this work, we developed an analytic model of networking energy consumption in a cloud computing environment. This analysis will serve as a basis for designing an energy-efficient architecture and related algorithms.

Simulating Energy Consumption of Wired Networks

Participants : Bogdan Florin Cornea, Anne-Cécile Orgerie.

Predicting the performance of applications, in terms of completion time and resource usage for instance, is critical to appropriately dimension resources that will be allocated to these applications. Current applications, such as web servers and Cloud services, require lots of computing and networking resources. Yet, these resource demands are highly fluctuating over time. Thus, adequately and dynamically dimension these resources is challenging and crucial to guarantee performance and cost-effectiveness. In the same manner, estimating the energy consumption of applications deployed over heterogeneous cloud resources is important in order to provision power resources and make use of renewable energies. Concerning the consumption of entire infrastructures, some studies show that computing resources represent the biggest part in Cloud’s consumption, while others show that, depending on the studied scenario, the energy cost of the network infrastructure that links the user to the computing resources can be bigger than the energy cost of the servers. 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. Through this tool, we have studied the energy consumption of data transfers in Clouds [19] . This work has been done in collaboration with the Avalon team from LIP in Lyon.

Resource allocation in a Cloud partially powered by renewable energy sources

Participants : Yunbo Li, Anne-Cécile Orgerie.

We propose here to design a disruptive approach to Cloud resource management which takes advantage of renewable energy availability to perform opportunistic tasks. To begin with, the considered Cloud is mono-site (i.e. all resources are in the same physical location) and performs tasks (like web hosting or MapReduce tasks) running in virtual machines. 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 needs to integrate a prediction model to be able to forecast these extra-power periods of time in order to schedule more work during these periods. Batteries will be used to guarantee that enough energy is available when switching on a new server working exclusively on renewable energy. Given a reliable prediction model, it is possible to design a scheduling algorithm that aims at optimizing resource utilization and energy usage, problem known to be NP-hard. The proposed heuristics will thus schedule tasks spatially (on the appropriate servers) and temporally (over time, with tasks that can be planed in the future).

This work is done in collaboration with Ascola team from LINA in Nantes.

SLA driven Cloud Auto-scaling for optimizing energy footprint

Participants : Sabbir Hasan Rochi, Jean-Louis Pazat.

As a direct consequence of the increasing popularity of Internet and Cloud Computing services, data centers are amazingly growing and hence have to urgently face energy consumption issues. At the Infrastructure-as-a-Service (IaaS) layer, Cloud Computing allows to dynamically adjust the provision of physical resources according to Platform-as-a-Service (PaaS) needs while optimizing energy efficiency of the data center.

The management of elastic resources in Clouds according to fluctuating workloads in the Software-as-a-Service (SaaS) applications and different Quality-of-Service (QoS) end-user’s expectations is a complex issue and cannot be done dynamically by a human intervention. We advocate the adoption of Autonomic Computing (AC) at each XaaS layer for responsiveness and autonomy in front of environment changes. At the SaaS layer, AC enables applications to react to a highly variable workload by dynamically adjusting the amount of resources in order to keep the QoS for the end users. Similarly, at the IaaS layer, AC enables the infrastructure to react to context changes by optimizing the allocation of resources and thereby reduce the costs related to energy consumption. However, problems may occur since those self-managed systems are related in some way (e.g. applications depend on services provided by a cloud infrastructure): decisions taken in isolation at given layer may interfere with other layers, leading whole system to undesired states.

We propose an approach driven by Service Level Agreements (SLAs) for Cloud auto-scaling. A SLA defines a formal contract between a service provider and a service consumer on an expected QoS level. The main idea of this thesis is to exploit the SLA requirements to (i) avoid the interferences between the Cloud autonomic managers by a cross-layer coordination of SLA contracts; (ii) fine-tune the resources needs according to SLA by proposing both dynamic resources provisioning for optimizing the energy footprint and dynamic reconfiguration at the SaaS level to optimize the expected QoS. In particular, we propose to address renewable energy in the SLA contract. The objective is twofold. First, for ecological reasons, it allows Cloud users to express their preferences about the energy provider and the nature of the energy in the data center. Then, for economic reasons, it takes advantage of renewable energy costs (expressed in the SLA) to reconfigure resource allocation and energy usage. The integration of such SLAs in each layer of the Cloud stack and their management by an autonomic manager or by the coordination of autonomic managers still remain open issues.

This work is done in collaboration with Ascola team from LINA in Nantes.

Simulating the impact of DVFS within SimGrid

Participants : Alexandra Carpen-Amarie, Christine Morin, Anne-Cécile Orgerie.

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. Furthermore, studying the impact of application configuration choices on energy is a difficult task, as not many platforms are equipped with the proper power measurement tools. The goal of this work is to enable energy-aware experimentations within the SimGrid simulation toolkit, by introducing a model of application energy consumption and enabling the use of Dynamic Voltage and Frequency Scaling (DVFS) techniques for the simulated platforms. We provide the methodology used to obtain accurate energy estimations, highlighting the simulator calibration phase. The proposed energy model is validated by means of a large set of experiments featuring several benchmarks and scientific applications. This work is available in the latest SimGrid release. This work is done in collaboration with the Mescal team from LIG in Grenoble.