Mateusz Jarus

Performance Evaluation and Energy Efficiency of High-Density HPC Platforms Based on Intel, AMD and ARM Processors

Due to growth of energy consumption by HPC servers and data centers many research efforts aim at addressing the problem of energy efficiency. Hence, the use of low power processors such as Intel Atom and ARM Cortex have recently gained more interest. In this article, we compare performance and energy efficiency of cutting-edge high-density HPC platform enclosures featuring either very high-performing processors such as Intel Core i7 or E7 yet having low power-efficiency, or the reverse i.e. energy efficient processors such as Intel Atom, AMD Fusion or ARM Cortex A9 yet with limited computing capacity. Our objective was to quantify in a very pragmatic way these general purpose CPUs using a set of reference benchmarks and applications run in an HPC environment, the trade-off that could exist between computing and power efficiency.

Modeling Data Center Building Blocks for Energy-Efficiency and Thermal Simulations

In this paper we present a concept and specification of Data Center Efficiency Building Blocks (DEBBs), which represent hardware components of a data center complemented by descriptions of their energy efficiency. Proposed building blocks contain hardware and thermodynamic models that can be applied to simulate a data center and to evaluate its energy efficiency. DEBBs are available in an open repository being built by the CoolEmAll project. In the paper we illustrate the concept by an example of DEBB defined for the RECS multi-server system including models of its power usage and thermodynamic properties. We also show how these models are affected by specific architecture of modeled hardware and differences between various classes of applications. Proposed models are verified by a comparison to measurements on a real infrastructure. Finally, we demonstrate how DEBBs are used in data center simulations.

Energy and Heat-Aware Metrics for Data Centers: Metrics Analysis in the Framework of CoolEmAll Project

CoolEmAll project aims at improving energy-efficiency of data centers. The main results of CoolEmAll include data center Simulation, Visualization and Decision tools (SVD Toolkit) and models of Data Center Efficiency Building Blocks (DEBBs). The resulting tools of CoolEmAll will permit planners and operators of data centers to carry out flexible and fast simulations to minimize the energy consumption on it and to reduce the associated greenhouse gas emissions. Several metrics have been proposed to assess the energy efficiency on data centers on the framework of CoolEmAll project. Unlike the common metrics on data center industry, the ones proposed in this project are focused not only on power consumption but also on dynamic heat-aware analysis. New metrics developed on CoolEmAll are (a) the Imbalance of Temperature of node, group of nodes and racks and (b) Rack Cooling Index adapted to a group of nodes. Node is defined as the smallest element of a data center to be modeled. This approach will permit to detect the cooling requirements and its source in order to implement strategies to reduce that energy demand. The paper describes the selected metrics and the results obtained on the CoolEmAll first prototype experimental tests.

CoolEmAll: Models and Tools for Planning and Operating Energy Efficient Data Centres

The need to improve how efficiently data centre operate is increasing due to the continued high demand for new data centre capacity combined with other factors such as the increased competition for energy resources. The financial crisis may have dampened data centre demand temporarily, but current projections indicate strong growth ahead. By 2020, it is estimated that annual investment in the construction of new data centres will rise to

Energy- and Heat-Aware HPC Benchmarks

50bn in the US, and

Top-Down Characterization Approximation based on performance counters architecture for AMD processors

220bn worldwide [23].

Energy Efficiency in HPC Data Centers: Latest Advances to Build the Path to Exascale

To evaluate data centers is tough. Several metrics are available to provide insight into their behaviour, but usually they are tested using simple benchmarks like LINPACK for HPC oriented data centers. A good choice of benchmarks is necessary to evaluate all the impact of applications on those data centers. One point that is often overlooked is their energy- and thermal-quality. To evaluate these qualities, adequate benchmarks are required from several points of view: from the nodes to the whole building. Classical benchmarks selection mainly focuses on time and raw performance. This article aims at shifting the focus towards an energy- and power-point of view. To this end, we select benchmarks able to evaluate data centers not only from this performance perspective, but also from the energy and thermal standpoint. We also provide insight into several classical benchmarks and method to select an adequate and small number of benchmarks in order to provide a sensible and minimum set of energy- and thermal-aware benchmarks for HPC systems.

Supporting supply process in charitable organizations by genetic algorithm

Abstract Due to the increasing complexity of the processors, developers often seek for tools that would simplify the process of finding bottlenecks while executing applications. Although more and more data may be collected from processors, usually much detailed knowledge about the internals of a given architecture is required to understand them. This paper introduces a Top-Down Characterization Approximation for the analysis of applications performance executed on AMD processors and is an extension of a Top-Down Method initially developed by Intel. Since not all required performance counters are available on AMD processors to calculate the exact values of metrics, this method was named as an approximation. It allows one to get a deeper understanding of different stages of program execution, compare different architectures and identify bottlenecks in out-of-order processors. It hides from the user the complexity of microarchitecture details and at the same time exposes the main contributors of inefficient program execution. This method aims at defining a few main metrics on top of performance counters to easily locate the main efficiency issues. At this time this method was applied to Intel processors only. The main reason behind it was the fact that it uses designated performance counters that are unique among different processors and its portability is not straightforward. Positive feedback from users encouraged the authors to develop a similar technique for AMD processors.

Gicomp and GreenOffice - Monitoring and Management Platforms for IT and Home Appliances

Nowadays, moderating energy consumption and building eco-friendly computing infrastructure is a major goal in large data centers. Moreover, data center energy usage has risen dramatically over the past decade and will continue to grow in-step with the High Performance Computing (HPC) intensive workloads which are at the heart of our modern life. The recent advances in the technology has driven the data center into a new phase of expansion featuring solutions with higher density. To this end, much has been done to increase server efficiency and IT space utilization. In this chapter, we will provide a state-of-the-art overview as regards energy-efficiency in High Performance Computing (HPC) facilities while describing the open challenges the research community has to face in the coming years to enable the building and usage of an Exascale platform by 2020.

Runtime power usage estimation of HPC servers for various classes of real-life applications

The problem arising e.g. in charitable organizations during supply process is analysed.The formal mathematical model and the lower bound for the criterion value is presented.The genetic algorithm and the specialized list heuristic approach are proposed.Extensive computational experiments for real world like data are reported.The algorithms were integrated with web interface into a decision supporting system. The paper concerns the optimization problem arising in charitable organizations during supply process. Such institutions are especially interested in minimizing the cost of purchase which consists of the prices at which particular products are bought as well as of the cost of their transportation. We present the formal mathematical model of the problem and the lower bound for the criterion value. We propose a genetic algorithm and the specialized list heuristic approach solving the case, which we prove is strongly NP-hard. The efficiency of implemented methods was checked in extensive computational experiments. The proposed algorithms have been integrated with the software system designed with a view of supporting charitable organizations.