On the potential of significance-driven execution for energy-aware HPC

Dynamic Voltage and Frequency Scaling (DVFS) exhibits fundamental limitations as a method to reduce energy consumption in computing systems. In the HPC domain, where performance is of highest priority and codes are heavily optimized to minimize idle time, DVFS has limited opportunity to achieve substantial energy savings. This paper explores if operating processors Near the transistor Threshold Volt- age (NTV) is a better alternative to DVFS for break- ing the power wall in HPC. NTV presents challenges, since it compromises both performance and reliability to reduce power consumption. We present a first of its kind study of a significance-driven execution paradigm that selectively uses NTV and algorithmic error tolerance to reduce energy consumption in performance- constrained HPC environments. Using an iterative algorithm as a use case, we present an adaptive execution scheme that switches between near-threshold execution on many cores and above-threshold execution on one core, as the computational significance of iterations in the algorithm evolves over time. Using this scheme on state-of-the-art hardware, we demonstrate energy savings ranging between 35% to 67%, while compromising neither correctness nor performance.

Significance Driven Computation: A Voltage-Scalable, Variation-Aware, Quality-Tuning Motion Estimator

In this paper we present a design methodology for algorithm/architecture co-design of a voltage-scalable, process variation aware motion estimator based on significance driven computation. The fundamental premise of our approach lies in the fact that all computations are not equally significant in shaping the output response of video systems. We use a statistical technique to intelligently identify these significant/not-so-significant computations at the algorithmic level and subsequently change the underlying architecture such that the significant computations are computed in an error free manner under voltage over-scaling. Furthermore, our design includes an adaptive quality compensation (AQC) block which "tunes" the algorithm and architecture depending on the magnitude of voltage over-scaling and severity of process variations. Simulation results show average power savings of similar to 33% for the proposed architecture when compared to conventional implementation in the 90 nm CMOS technology. The maximum output quality loss in terms of Peak Signal to Noise Ratio (PSNR) was similar to 1 dB without incurring any throughput penalty.

The CACTOS Vision of Context-Aware Cloud Topology Optimization and Simulation

Recent advances in hardware development coupled with the rapid adoption and broad applicability of cloud computing have introduced widespread heterogeneity in data centers, significantly complicating the management of cloud applications and data center resources. This paper presents the CACTOS approach to cloud infrastructure automation and optimization, which addresses heterogeneity through a combination of in-depth analysis of application behavior with insights from commercial cloud providers. The aim of the approach is threefold: to model applications and data center resources, to simulate applications and resources for planning and operation, and to optimize application deployment and resource use in an autonomic manner. The approach is based on case studies from the areas of business analytics, enterprise applications, and scientific computing.

A Significance-Driven Programming Framework for Energy-Constrained Approximate Computing

Approximate execution is a viable technique for energy-con\-strained environments, provided that applications have the mechanisms to produce outputs of the highest possible quality within the given energy budget.
We introduce a framework for energy-constrained execution with controlled and graceful quality loss. A simple programming model allows users to express the relative importance of computations for the quality of the end result, as well as minimum quality requirements. The significance-aware runtime system uses an application-specific analytical energy model to identify the degree of concurrency and approximation that maximizes quality while meeting user-specified energy constraints. Evaluation on a dual-socket 8-core server shows that the proposed
framework predicts the optimal configuration with high accuracy, enabling energy-constrained executions that result in significantly higher quality compared to loop perforation, a compiler approximation technique.

Exploiting Significance of Computations for Energy-Constrained Approximate Computing

Approximate execution is a viable technique for environments with energy constraints, provided that applications are given the mechanisms to produce outputs of the highest possible quality within the available energy budget. This paper introduces a framework for energy-constrained execution with controlled and graceful quality loss. A simple programming model allows developers to structure the computation in different tasks, and to express the relative importance of these tasks for the quality of the end result. For non-significant tasks, the developer can also supply less costly, approximate versions. The target energy consumption for a given execution is specified when the application is launched. A significance-aware runtime system employs an application-specific analytical energy model to decide how many cores to use for the execution, the operating frequency for these cores, as well as the degree of task approximation, so as to maximize the quality of the output while meeting the user-specified energy constraints. Evaluation on a dual-socket 16-core Intel platform using 9 benchmark kernels shows that the proposed framework picks the optimal configuration with high accuracy. Also, a comparison with loop perforation (a well-known compile-time approximation technique), shows that the proposed framework results in significantly higher quality for the same energy budget.

Inexact-aware architecture design for ultra-low power bio-signal analysis

This study introduces an inexact, but ultra-low power, computing architecture devoted to the embedded analysis of bio-signals. The platform operates at extremely low voltage supply levels to minimise energy consumption. In this scenario, the reliability of static RAM (SRAM) memories cannot be guaranteed when using conventional 6-transistor implementations. While error correction codes and dedicated SRAM implementations can ensure correct operations in this near-threshold regime, they incur in significant area and energy overheads, and should therefore be employed judiciously. Herein, the authors propose a novel scheme to design inexact computing architectures that selectively protects memory regions based on their significance, i.e. their impact on the end-to-end quality of service, as dictated by the bio-signal application characteristics. The authors illustrate their scheme on an industrial benchmark application performing the power spectrum analysis of electrocardiograms. Experimental evidence showcases that a significance-based memory protection approach leads to a small degradation in the output quality with respect to an exact implementation, while resulting in substantial energy gains, both in the memory and the processing subsystem.

Energy-aware Workflow Scheduling and Fuzzy Logic Based Capacity Provisioning in Cloud Environment

Thesis (Master's)--University of Washington, 2015

Dynamic power- and failure-aware cloud resources allocation for Sets of independent tasks

Cloud computing is increasingly being adopted in different scenarios, like social networking, business applications, scientific experiments, etc. Relying in virtualization technology, the construction of these computing environments targets improvements in the infrastructure, such as power-efficiency and fulfillment of users’ SLA specifications. The methodology usually applied is packing all the virtual machines on the proper physical servers. However, failure occurrences in these networked computing systems can induce substantial negative impact on system performance, deviating the system from ours initial objectives. In this work, we propose adapted algorithms to dynamically map virtual machines to physical hosts, in order to improve cloud infrastructure power-efficiency, with low impact on users’ required performance. Our decision making algorithms leverage proactive fault-tolerance techniques to deal with systems failures, allied with virtual machine technology to share nodes resources in an accurately and controlled manner. The results indicate that our algorithms perform better targeting power-efficiency and SLA fulfillment, in face of cloud infrastructure failures.

SPARTS: simulator for power aware and real-time systems

Real-time systems demand guaranteed and predictable run-time behaviour in order to ensure that no task has missed its deadline. Over the years we are witnessing an ever increasing demand for functionality enhancements in the embedded real-time systems. Along with the functionalities, the design itself grows more complex. Posed constraints, such as energy consumption, time, and space bounds, also require attention and proper handling. Additionally, efficient scheduling algorithms, as proven through analyses and simulations, often impose requirements that have significant run-time cost, specially in the context of multi-core systems. In order to further investigate the behaviour of such systems to quantify and compare these overheads involved, we have developed the SPARTS, a simulator of a generic embedded real- time device. The tasks in the simulator are described by externally visible parameters (e.g. minimum inter-arrival, sporadicity, WCET, BCET, etc.), rather than the code of the tasks. While our current implementation is primarily focused on our immediate needs in the area of power-aware scheduling, it is designed to be extensible to accommodate different task properties, scheduling algorithms and/or hardware models for the application in wide variety of simulations. The source code of the SPARTS is available for download at [1].

Job phasing aware preemption deferral

Preemptions account for a non-negligible overhead during system execution. There has been substantial amount of research on estimating the delay incurred due to the loss of working sets in the processor state (caches, registers, TLBs) and some on avoiding preemptions, or limiting the preemption cost. We present an algorithm to reduce preemptions by further delaying the start of execution of high priority tasks in fixed priority scheduling. Our approaches take advantage of the floating non-preemptive regions model and exploit the fact that, during the schedule, the relative task phasing will differ from the worst-case scenario in terms of admissible preemption deferral. Furthermore, approximations to reduce the complexity of the proposed approach are presented. Substantial set of experiments demonstrate that the approach and approximations improve over existing work, in particular for the case of high utilisation systems, where savings of up to 22% on the number of preemption are attained.

Handling mobility on a QoS-Aware servicebased framework for mobile systems

Mobile applications are becoming increasingly more complex and making heavier demands on local system resources. Moreover, mobile systems are nowadays more open, allowing users to add more and more applications, including third-party developed ones. In this perspective, it is increasingly expected that users will want to execute in their devices applications which supersede currently available resources. It is therefore important to provide frameworks which allow applications to benefit from resources available on other nodes, capable of migrating some or all of its services to other nodes, depending on the user needs. These requirements are even more stringent when users want to execute Quality of Service (QoS) aware applications, such as voice or video. The required resources to guarantee the QoS levels demanded by an application can vary with time, and consequently, applications should be able to reconfigure themselves. This paper proposes a QoS-aware service-based framework able to support distributed, migration-capable, QoS-enabled applications on top of the Android Operating system.

Time-bounded distributed QoS-aware service configuration in heterogeneous cooperative environments

The scarcity and diversity of resources among the devices of heterogeneous computing environments may affect their ability to perform services with specific Quality of Service constraints, particularly in dynamic distributed environments where the characteristics of the computational load cannot always be predicted in advance. Our work addresses this problem by allowing resource constrained devices to cooperate with more powerful neighbour nodes, opportunistically taking advantage of global distributed resources and processing power. Rather than assuming that the dynamic configuration of this cooperative service executes until it computes its optimal output, the paper proposes an anytime approach that has the ability to tradeoff deliberation time for the quality of the solution. Extensive simulations demonstrate that the proposed anytime algorithms are able to quickly find a good initial solution and effectively optimise the rate at which the quality of the current solution improves at each iteration, with an overhead that can be considered negligible.