Georges Da Costa

The GREEN-NET framework: Energy efficiency in large scale distributed systems

The question of energy savings has been a matter of concern since a long time in the mobile distributed systems and battery-constrained systems. However, for large-scale non-mobile distributed systems, which nowadays reach impressive sizes, the energy dimension (electrical consumption) just starts to be taken into account. In this paper, we present the GREEN-NET1 framework which is based on 3 main components: an ON/OFF model based on an Energy Aware Resource Infrastructure (EARI), an adapted Resource Management System (OAR) for energy efficiency and a trust delegation component to assume network presence of sleeping nodes.

Energy-aware simulation with DVFS

In recent years, research has been conducted in the area of large systems models, especially distributed systems, to analyze and understand their behavior. Simulators are now commonly used in this area and are becoming more complex. Most of them provide frameworks for simulating application scheduling in various Grid infrastructures, others are specifically developed for modeling networks, but only a few of them simulate energy-efficient algorithms. This article describes which tools need to be implemented in a simulator in order to support energy-aware experimentation. The emphasis is on DVFS simulation, from its implementation in the simulator CloudSim to the whole methodology adopted to validate its functioning. In addition, a scientific application is used as a use case in both experiments and simulations, where the close relationship between DVFS efficiency and hardware architecture is highlighted. A second use case using Cloud applications represented by DAGs, which is also a new functionality of CloudSim, demonstrates that the DVFS efficiency also depends on the intrinsic middleware behavior.

Energy-aware service allocation

In this paper we study the problem of energy-aware resource allocation for hosting long-term services or on-demand computing jobs in clusters, e.g., deployed as part of computing infrastructures. We formalize the problem as three constrained optimization problems: maximize job performance under power consumption constraints, minimize power consumption under job performance constraints, and optimize a linear combination of power consumption and job performance. These problems are NP-hard but, given an instance, a bound on the optimal solution can be computed via a rational linear program. We propose polynomial heuristics for all three problems. Simulation experiments show that in all three cases some heuristics can achieve results close to optimal, i.e., lead to good job performance while conserving energy.

Energy Consumption of Residential and Professional Switches

Precise evaluation of network appliance energy consumption is necessary to accurately model or simulate the power consumption of distributed systems. In this paper we evaluate the influence of traffic onto the consumption of electrical power of four switches found in home and professional environments. First we describe our measurement and data analysis approach, and how our results can be used for estimating the power consumption when knowing the average traffic bandwidth.Then we present the measurement results of two residential switches, and two professional switches. For each type we present regression models and parameters describing their quality. Similar to other works we find that for one of the switches the power consumption actually drops for high traffic loads, while for the others the situation is reverse. Measures justify that during most energy consumption evaluation, network appliance energy cost can be approximated as constant. This work gives information on the possible changes of this cost.

Methodology of measurement for energy consumption of applications

For IT systems, energy awareness can be improved in two ways, (i) in a static or (ii) in a dynamic way. The first way leads to building energy efficient hardware that runs fast and consumes only a few watts. The second way consists of reacting to instantaneous power consumption, and of taking decisions that will reduce this consumption.

A Runtime Framework for Energy Efficient HPC Systems without a Priori Knowledge of Applications

The rising computing demands of scientific endeavors often require the creation and management of High Performance Computing (HPC) systems for running experiments and processing vast amounts of data. These HPC systems generally operate at peak performance, consuming a large quantity of electricity, even though their workload varies over time. Understanding the behavioral patterns (i.e., phases) of HPC systems during their use is key to adjust performance to resource demand and hence improve the energy efficiency. In this paper, we describe (i) a method to detect phases of an HPC system based on its workload, and (ii) a partial phase recognition technique that works cooperatively with on-the-fly dynamic management. We implement a prototype that guides the use of energy saving capabilities to demonstrate the benefits of our approach. Experimental results reveal the effectiveness of the phase detection method under real-life workload and benchmarks. A comparison with baseline unmanaged execution shows that the partial phase recognition technique saves up to 15% of energy with less than 1% performance degradation.

Energy-efficient, thermal-aware modeling and simulation of data centers

This paper describes the CoolEmAll project and its approach for modeling and simulating energy-efficient and thermal-aware data centers. The aim of the project was to address energy-thermal efficiency of data centers by combining the optimization of IT, cooling and workload management. This paper provides a complete data center model considering the workload profiles, the applications profiling, the power model and a cooling model. Different energy efficiency metrics are proposed and various resource management and scheduling policies are presented. The proposed strategies are validated through simulation at different levels of a data center.

Modeling and simulation of data center energy-efficiency in coolemall

In this paper we present an overview of the CoolEmAll project which addresses the important problem of data center energy efficiency. To this end, CoolEmAll aims at delivering advanced simulation, visualization and decision support tools along with open models of data center building blocks to be used in simulations. Both building blocks and the toolkit will take into account aspects that have major impact on actual energy consumption such as cooling solutions, properties of applications, and workload and resource management policies. In the paper we describe the CoolEmAll approach, its expected results and an environment for their verification.

Energy-aware resource allocation

This paper deals with the reduction of energy consumption in large scale systems, especially by taking into account the impact of energy consumption for server consolidation. Decreasing the number of physical hosts used while ensuring a certain level of quality of services is the goal of our approach. We introduce a metric called energetic yield which represents the quality of a task placement on a subset of machines, while taking into account quality of service and energy efficiency aspects. It measures the difference between resources required by a job and what the system allocates ultimately, while trying to save energy. Our work aims at minimizing this difference. We propose placement heuristics that are compared to the optimal solution and to a related system. In this paper, we present a set of experiments showing the relevance of this metric in order to reduce significantly energy consumption.

DNA-inspired scheme for building the energy profile of HPC systems

Energy usage is becoming a challenge for the design of next generation large scale distributed systems. This paper explores an innovative approach of profiling such systems. It proposes a DNA-like solution without making any assumptions on the running applications and used hardware. This profiling based on internal counters usage and energy monitoring allows to isolate specific phases during the execution and enables some energy consumption control and energy usage prediction. First experimental validations of the system modeling are presented and analyzed.