Large language models (LLMs) have become powerful tools for automated code generation. Yet, they remain prone to both syntax and logic errors that limit their effectiveness in real-world software development. This paper comprehensively evaluates LLM-generated code utilizing the Codeforces Benchmark, and proposes XML-Structured Prompts (XSP) to improve code quality. Specifically, we present three creative strategies: XSP for improving problem comprehension, Syntax-Aware XSP (SA-XSP) for mitigating syntax errors, and Testcase-Driven Logic Error Debugging (TDLED) for resolving logic errors through test case feedback. Our experiments on 30 LLMs, ranging from$0.5 ~\mathrm{B}, 70 ~\mathrm{B}$, to 671 B parameters and including 5 reasoning models, demonstrate that the proposed methods improve accuracy by up to 27.5 %, especially in smaller models, with syntax error reductions of up to 60 % and logic error corrections by up to 30%.

Recently, Large Language Models (LLMs) have made substantial progress in code generation, but they still frequently generate code containing logic errors or syntax bugs. While research has focused on improving performance through fine-tuning and data collection, less attention has been given to analyzing error patterns and employing prompt-engineering to address these issues. Existing benchmarks primarily assess LLMs on easy to intermediate-level coding tasks, often neglecting more complex challenges involving advanced algorithms and data structures. Additionally, data contamination in these benchmarks limits their ability to accurately measure the capability of LLMs in code generation. In this paper, we present the new USACO Benchmark, derived from the USA Computing Olympiad (USACO) competition, to evaluate 11 closed and open-source LLMs. Through a detailed analysis, we identify common code generation errors across the models and propose Hint-Driven Prompts to address logic errors, alongside the Syntax Mitigation Prompt to reduce syntax bugs. Our results demonstrate that the Hint-Driven Prompt boosts pass rates for DBRX 132B, Deepseek-Coder 33B, Codegemma 7B, Codellama 7B, Llama 3, and GPT-4o by 6.6×, 4.7×, 3×, 2.5×, 2.1×, and 25%, respectively. Additionally, the Syntax Mitigation Prompt significantly reduces syntax errors, with reductions of 71.32% for Codegemma 7B, 25.56% for Deepseek-Coder 33B, 23.39% for Llama 3, and 11.19% for Codellama 70B.

The significance of accurate power measurement in cloud computing can be attributed to its increasing usability as well as for attaining energy savings through optimization and advancing green computing in diverse cloud platforms like AWS, GCP, Chameleon Cloud, and Microsoft Azure. The challenge of providing a single yet accurate methodology, suitable for the cloud environment, is exacerbated by factors like numerous heterogeneity of the cloud, unavailability of administrative access required by precise methods like Intel's Running Average Power Limit (RAPL) and the inaccuracy of widely-accepted methods like Thermal Design Power (TDP) based estimation. In this paper, we propose Cloud Power Meter (CPM) in pursuit of sustainability in cloud operations, which integrates a decision tree with multi-variable polynomial regression. Being an ML-based method that is trained using diverse cloud instances from Amazon, Microsoft, and Google, CPM is capable of providing precise power estimation on the cloud platforms without administrative privileges. The decision tree guarantees performance capacity across heterogeneous cloud environments by locating the “CPM Instance” - the closest match of instances that are not familiar to CPM. In evaluation through the Standard Performance Evaluation Corporation (SPEC) benchmark's CPU power consumption, CPM outperforms 14 existing power models and its power estimation matches Intel's RAPL power model. To evaluate the effectiveness of CPM, we conduct an extensive analysis of approximately 2.6 million real-world Virtual Machines (VMs) running on Microsoft Azure. CPM successfully identifies opportunities to save over 80,000 kilowatt-hours (kWh) of energy in a single month through optimized resource allocation. Notably, about 25,000 kWh of this energy is being wasted by more than 17% of the VMs. Further, CPM's energy estimations reveal that 366 VMs (out of roughly 2.6 million) wasted nearly 15,000 kWh of energy despite never exceeding 2% CPU utilization.

Large Language Models (LLMs), such as GPT-3, ChatGPT, and GPT-4, have demonstrated enormous potential across a range of tasks and attracted over 100 million users globally in recent months. However, these LLMs are resource-intensive and contribute significantly to carbon emissions. Currently, our understanding of their carbon impact remains insufficient due to the lack of reliable measurement tools, standard methodologies, and evaluation metrics. To bridge this gap, this paper conducts a thorough study on the carbon impact of various open-source LLMs, including GPT-J 6B, GPT Neo 2.7B, GPTNEO 1.3B, and GPT-2 at the inference stage, utilizing the Software Carbon Intensity (SCI) specification released by the Green Software Foundation. The primary contributions of our research are: (1) We propose a quantitative framework that measures and contrasts the environmental impacts of different LLMs; (2) We illustrate that high-carbon LLMs do not necessarily provide superior model quality than their low-carbon counterparts; and (3) We find that the carbon emissions are primarily driven by embodied carbon in LLMs and that employing GPUs, as opposed to CPUs, can substantially reduce carbon emissions.

Cloud computing has grown at an unprecedented rate in recent years, which leads to two new significant challenges. First, cloud expenditure is increasing rapidly and reducing cloud expenditure has become the top priority of enterprises relying on cloud services. Second, cloud computing consumes significant amounts of energy and generates massive carbon emissions. This paper presents a comprehensive study to reduce the cost and carbon emissions of cloud computing. Through analyzing over 2.6 million real-world Azure virtual machines (VMs) and carbon intensity data provided by WattTime, it proposes a metric to quantify cloud waste and trains a machine learning model to estimate the power consumption and carbon emissions of VMs. Furthermore, it proposes two VM scheduling algorithms, Reschedule by Threshold (RT) and Reschedule by Averages (RA), to reduce the carbon emissions of cloud workloads, and the other two algorithms, the Shutdown (SD) algorithm and the Core Reduction (CR) algorithm, to reduce both cloud cost and carbon emissions. The simulation results running on the 2019 Azure VM trace demonstrate that the proposed algorithms could reduce nearly 3.5 million pounds of CO2 emissions and help cloud users save approximately ${\$}$13.8 million dollars per month.

With the remarkable success of deep learning recently, efficient network compression algorithms are urgently demanded for releasing the potential computational power of edge devices, such as smartphones or tablets. However, optimal network pruning is a non-trivial task which mathematically is an NP-hard problem. Previous researchers explain training a pruned network as buying a lottery ticket. In this paper, we investigate the Magnitude-Based Pruning (MBP) scheme and analyze it from a novel perspective through Fourier analysis on the deep learning model to guide model designation. Besides explaining the generalization ability of MBP using Fourier transform, we also propose a novel two-stage pruning approach, where one stage is to obtain the topological structure of the pruned network and the other stage is to retrain the pruned network to recover the capacity using knowledge distillation from lower to higher on the frequency domain. Extensive experiments on CIFAR-10 and CIFAR-100 demonstrate the superiority of our novel Fourier analysis based MBP compared to other traditional MBP algorithms.

In recent years, computational accessibility has enabled the use of Deep Neural Network (DNN) for computer vision applications on devices with limited computational resources. We focus on the real-time object detection algorithms deployed on UAV -friendly devices. The hardware deployed on UAV must be lightweight and thus limited in processing power, memory, and storage capacity. Lightweight modeling architecture does not suffice for high-recall reconnaissance applications. In this paper, we propose to reduce power consumption of YOLOv5 DNN architecture. We decided to use compressed convolutional technique, transfer learning, backbone shrinkage, and scale prediction to reduce the number of learnable parameters from the YOLOv5model. Our approach reduced the size of the model significantly and lowered the power consumption in turn. GPU memory and the Billion Floating-Point Operations Per Second (GFLOPS) for the YOLOv5model will keep the performance measure of the model as the baseline state-of-the-art. The best resulting model has a 63.86% mean average precision (mAP) and a GFLOPS of 97.7 on “DIOR”, an overhead imagery data set. The proposed approach has lowered GPU memory consumption of the model by 34% and lowered the energy consumption by 10 Watts compared to the baseline model.

Algorithmic bias is of increasing concern, both to the research community, and society at large. Bias in AI is more abstract and unintuitive than traditional forms of discrimination and can be more difficult to detect and mitigate. A clear gap exists in the current literature on evaluating the relative bias in the performance of multi-class classifiers. In this work, we propose two simple yet effective metrics, Combined Error Variance (CEV) and Symmetric Distance Error (SDE), to quantitatively evaluate the class-wise bias of two models in comparison to one another. By evaluating the performance of these new metrics and by demonstrating their practical application, we show that they can be used to measure fairness as well as bias. These demonstrations show that our metrics can address specific needs for measuring bias in multi-class classification. Demonstration code is available at https://github.com/gentry-atkinson/CEV_SDE_demo.git.

Knowledge distillation has become one of the most important model compression techniques by distilling knowledge from larger teacher networks to smaller student ones. Although great success has been achieved by prior distillation methods via delicately designing various types of knowledge, they overlook the functional properties of neural networks, which makes the process of applying those techniques to new tasks unreliable and non-trivial. To alleviate such problem, in this paper, we initially leverage Lipschitz continuity to better represent the functional characteristic of neural networks and guide the knowledge distillation process. In particular, we propose a novel Lipschitz Continuity Guided Knowledge Distillation framework to faithfully distill knowledge by minimizing the distance between two neural networks’ Lipschitz constants, which enables teacher networks to better regularize student networks and improve the corresponding performance. We derive an explainable approximation algorithm with an explicit theoretical derivation to address the NP-hard problem of calculating the Lipschitz constant. Experimental results have shown that our method outperforms other benchmarks over several knowledge distillation tasks (e.g., classification, segmentation and object detection) on CIFAR-100, ImageNet, and PASCAL VOC datasets. Our code is available at https://github.com/42Shawn/LONDON/tree/master.

Only recently has the software engineering community started conducting research on developing energy efficient software, or green software. This is shadowed when compared to the research already produced in the computer hardware community. While research in green software is rapidly increasing, several recent studies with software engineers show that they still miss techniques, knowledge, and tools to develop greener software. Indeed, all such studies suggest that green software should be part of a modern Computer Science Curriculum.

For decades, the x86 architecture supported by Intel and AMD has been the dominate target for software development. Recently, ARM has solidified itself as a highly competitive and promising CPU architecture by exhibiting both high performance and low power consumption simultaneously. In the foreseeable future, a copious amount of software will be fully migrated to the ARM architecture or support both x86 and ARM simultaneously. Nevertheless, software ports from x86 to ARM are not trivial for a number of reasons. First, it is time consuming to write code that resolves all compatibility issues for a new architecture. Second, specific hardware (e.g. ARM chips) and supporting toolkits (e.g. libraries and compilers) may not be readily available for developers, which will delay the porting process. Third, it is hard to predict the performance of software before testing it on production chips. In this paper, we strive to tackle these challenges by proposing an instruction prediction method that can automatically generate AARCH64 code from existing x86-64 executables. Although the generated code might not be directly executable, it provides a cheap and efficient solution for developers to estimate certain runtime metrics before actually building, deploying and testing code on an ARM-based CPU. Our experimental results show that AARCH64 instructions derived using prediction can achieve a high Bilingual Evaluation Understudy (BLEU) Score. This indicates a quality match between generated executables and natively ported AARCH64 software.

Mobile apps such as games and virtual reality(VR) are getting increasingly popular but they drain battery quickly due to the heavy graphics rending process. Currently, Open Graphics Library for Embedded Systems (OpenGL ES) is the dominating API for rendering advanced graphics on embedded and mobile systems. Despite the attracting usability of OpenGL ES, the lacking support of multi-threading limits its performance and power efficiency on modern multicore mobile chips, especially when the big.LITTLE architecture has become the de facto industry standard of mobile phones. Vulkan was recently proposed to address the weaknesses of OpenGL but its performance and energy efficiency on the big.LITTLE architecture has not been fully explored yet. This paper conducts a comprehensive study to compare the performance and energy efficiency of Vulkan versus OpenGL ES on an ARM processor with both high performance cores (i.e. big cores) and low power cores (i.e. LITTLE cores). Our experimental results show that 1) Vulkan can save up to 24% of energy by leveraging multi-threading and parallel execution on LITTLE cores for heavy workloads; and 2) Vulkan can render at a much higher frame rate when OpenGL ES has reached its full capability. Meanwhile, writing efficient Vulkan code is not trivial and the performance/energy gains are negligible for light workloads. The clear tradeoff between optimizing verbose Vulkan code manually and potential performance or energy efficiency benefits should be carefully considered.

The major challenge in audio-visual event localization task lies in how to fuse information from multiple modalities effectively. Recent works have shown that the attention mechanism is beneficial to the fusion process. In this paper, we propose a novel joint attention mechanism with multi-modal fusion methods for audio-visual event localization. Particularly, we present a concise yet valid architecture that effectively learns representations from multiple modalities in a joint manner. Initially, visual features are combined with auditory features and then turned into joint representations. Next, we make use of the joint representations to attend to visual features and auditory features, respectively. With the help of this joint co-attention, new visual and auditory features are produced, and thus both features can enjoy the mutually improved benefits from each other. It is worth noting that the joint co-attention unit is recursive meaning that it can be performed multiple times for obtaining better joint representations progressively. Extensive experiments on the public AVE dataset have shown that the proposed method achieves significantly better results than the state-of-the-art methods.

Deep neural networks (DNNs) have been extremely successful in solving many challenging AI tasks in natural language processing, speech recognition, and computer vision nowadays. However, DNNs are typically computation intensive, memory demanding, and power hungry, which significantly limits their usage on platforms with constrained resources. Therefore, a variety of compression techniques (e.g., quantization, pruning, and knowledge distillation) have been proposed to reduce the size and power consumption of DNNs. Blockwise knowledge distillation is one of the compression techniques that can effectively reduce the size of a highly complex DNN. However, it is not widely adopted due to its long training time. In this article, we propose a novel parallel blockwise distillation algorithm to accelerate the distillation process of sophisticated DNNs. Our algorithm leverages local information to conduct independent blockwise distillation, utilizes depthwise separable layers as the efficient replacement block architecture, and properly addresses limiting factors (e.g., dependency, synchronization, and load balancing) that affect parallelism. The experimental results running on an AMD server with four Geforce RTX 2080Ti GPUs show that our algorithm can achieve 3x speedup plus 19 percent energy savings on VGG distillation, and 3.5x speedup plus 29 percent energy savings on ResNet distillation, both with negligible accuracy loss. The speedup of ResNet distillation can be further improved to 3.87 when using four RTX6000 GPUs in a distributed cluster.

Recently, there has been a growing interest in wearable sensors which provides new research perspectives for 360 ° video analysis. However, the lack of 360 ° datasets in literature hinders the research in this field. To bridge this gap, in this paper we propose a novel Egocentric (first-person) 360° Kinetic human activity video dataset (EgoK360). The EgoK360 dataset contains annotations of human activity with different sub-actions, e.g., activity Ping-Pong with four sub-actions which are pickup-ball, hit, bounce-ball and serve. To the best of our knowledge, EgoK360 is the first dataset in the domain of first-person activity recognition with a 360° environmental setup, which will facilitate the egocentric 360 ° video understanding. We provide experimental results and comprehensive analysis of variants of the two-stream network for 360 egocentric activity recognition. The EgoK360 dataset can be downloaded from https://egok360.github.io/.

Nowadays 360 video analysis has become a significant research topic in the field since the appearance of high-quality and low-cost 360 wearable devices. In this paper, we propose a novel LiteFlowNet360 architecture for 360 videos optical flow estimation. We design LiteFlowNet360 as a domain adaptation framework from perspective video domain to 360 video domain. We adapt it from simple kernel transformation techniques inspired by Kernel Transformer Network (KTN) to cope with inherent distortion in 360 videos caused by the sphere-to-plane projection. First, we apply an incremental transformation of convolution layers in feature pyramid network and show that further transformation in inference and regularization layers are not important, hence reducing the network growth in terms of size and computation cost. Second, we refine the network by training with augmented data in a supervised manner. We perform data augmentation by projecting the images in a sphere and re-projecting to a plane. Third, we train LiteFlowNet360 in a self-supervised manner using target domain 360 videos. Experimental results show the promising results of 360 video optical flow estimation using the proposed novel architecture.

Unstructured neural network pruning is an effective technique that can significantly reduce theoretical model size, computation demand and energy consumption of large neural networks without compromising accuracy. However, a number of fundamental questions about pruning are not answered yet. For example, do the pruned neural networks contain the same representations as the original network? Is pruning a compression or evolution process? Does pruning only work on trained neural networks? What is the role and value of the uncovered sparsity structure? In this paper, we strive to answer these questions by analyzing three unstructured pruning methods (magnitude based pruning, post-pruning re-initialization, and random sparse initialization). We conduct extensive experiments using the Singular Vector Canonical Correlation Analysis (SVCCA) tool to study and contrast layer representations of pruned and original ResNet, VGG, and ConvNet models. We have several interesting observations: 1) Pruned neural network models evolve to substantially different representations while still maintaining similar accuracy. 2) Initialized sparse models can achieve reasonably good accuracy compared to well-engineered pruning methods. 3) Sparsity structures discovered by pruning models are not inherently important or useful.

Hadoop MapReduce is a leading open source framework that supports the realization of the Big Data revolution and serves as a pioneering platform in ultra large amount of information storing and processing. However, tuning a MapReduce system has become a difficult task because a large number of parameters restrict its performance, many of which are related with shuffle, a complicated phase between map and reduce functions, including sorting, grouping, and HTTP transferring. During shuffle phase, a large mount of time is spent on disk I/O due to the low speed of data throughput. In this paper, we build a mathematical model to judge the computing complexity of different operating orders within map-side shuffle, so that a faster execution can be achieved through reconfiguring the order of sorting and grouping. Furthermore, a three-dimensional exploring space of the performance is expanded, with which, some sampled features during shuffle stage, such as key number, spilling file number, and the variances of intermediate results, are collected to support the evaluation of computing complexity of each operating order. Thus, an optimized reconfiguration of map-side shuffle architecture can be achieved within Hadoop without extra disk I/O induced. Comparing with the original Hadoop implementation, the results show that our reconfigurable architecture gains up to 2.37χ speedup to finish the map-side shuffle work.

Green cloud is an emerging new technology in the computing world in which memory is a critical component. Phase-change memory (PCM) is one of the most promising alternative techniques to the dynamic random access memory (DRAM) that faces the scalability wall. Recent research has been focusing on the multi-level cell (MLC) of PCM. By precisely arranging multiple levels of resistance inside a PCM cell, more than one bit of data can be stored in one single PCM cell. However, the MLC PCM suffers from the degradation of performance compared to the single-level cell(SLC) PCM, due to the longer memory access time. In this paper, we present a genetic-based optimization algorithm for chip multiprocessor (CMP) equipped with PCM memory in green clouds. The proposed genetic-based algorithm not only schedules and assigns tasks to cores in the CMP system, but also provides a PCM MLC configuration that balances the PCM memory performance as well as the efficiency. The experimental results show that our genetic-based algorithm can significantly reduce the maximum memory usage by 76.8 percent comparing with the uniform SLC configuration, and improve the efficiency of memory usage by 127 percent comparing with the uniform 4 bits/cell MLC configuration. Moreover, the performance of the system is also improved by 24.5 percent comparing with the uniform 4 bits/cell MLC configuration in terms of total execution time.

Accelerator-based heterogeneous systems can provide high performance and energy efficiency, both of which are key design goals in high performance computing. To fully realize the potential of heterogeneous architectures, software must optimally exploit the hosts' and accelerators' processing and power-saving capabilities. Yet, previous studies mainly focus on using hosts and accelerators to boost application performance. Power-saving features to improve the energy efficiency of parallel programs, such as Dynamic Voltage and Frequency Scaling (DVFS), remain largely unexplored. Recognizing that energy efficiency is a different objective than performance and should therefore be independently pursued, we study how to judiciously distribute computation between hosts and accelerators for energy optimization. We further explore energy-saving scheduling in combination with computation distribution for even larger gains. Moreover, we present PEACH, an analytical model for Performance and Energy Aware Cooperative Hybrid computing. With just a few system- and application-dependent parameters, PEACH accurately captures the performance and energy impact of computation distribution and energy-saving scheduling to quickly identify the optimal coupled strategy for achieving the best performance or the lowest energy consumption. PEACH thus eliminates the need for extensive profiling and measurement. Experimental results from two GPU-accelerated heterogeneous systems show that PEACH predicts the performance and energy of the studied codes with less than 3% error and successfully identifies the optimal strategy for a given objective.

Boosting performance and energy efficiency of scientific applications running on high performance computing systems arise cruicially nowadays. Software and hardware based solutions for improving communication performance have been recognized as significant means of achieving performance gain and thus energy savings for such applications. As a fundamental component of most numerical linear algebra algorithms, improving performance and energy efficiency of distributed matrix multiplication is of major concerns. For such purposes, we propose a high performance communication scheme that fully exploits network bandwidth via non-blocking pipeline broadcast with tuned chunk size. Empirically, substantial performance gain up to 8.4% and energy savings up to 6.9% are achieved compared to blocking pipeline broadcast, and against binomial tree broadcast, performance gain up to 6.5% and energy savings up to 6.1% are observed on a 64-core cluster.

Improving energy efficiency is an ongoing challenge in HPC because of the ever-increasing need for performance coupled with power and economic constraints. Though GPU-accelerated heterogeneous computing systems are capable of delivering impressive performance, it is necessary to explore all available power-aware technologies to meet the inevitable energy efficiency challenge. In this paper, we experimentally study the impacts of DVFS on application performance and energy efficiency for GPU computing and compare them with those of DVFS for CPU computing. Based on a power-aware heterogeneous system that includes dual Intel Sandy Bridge CPUs and the latest Nvidia K20c Kepler GPU, the study provides numerous new insights, general trends and exceptions of DVFS for GPU computing. In general, the effects of DVFS on a GPU differ from those of DVFS on a CPU. For example, on a GPU running compute-bound high-performance and high-throughput workloads, the system performance and the power consumption are approximately proportional to the GPU frequency. Hence, with a permissible power limit, increasing the GPU frequency leads to better performance without incurring a noticeable increase in energy. This paper further provides detailed analytical explanations of the causes of the observed trends and exceptions. The findings presented in this paper have the potential to impact future CPU and GPU architectures to achieve better energy efficiency and point out directions for designing effective DVFS schedulers for heterogeneous systems.

Many high performance large-scale storage systems will experience significant workload increases as their user base and content availability grow over time. The U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) center hosts one such system that has recently undergone a period of rapid growth as its user population grew nearly 400% in just about three years. When administrators of these massive storage systems face the challenge of meeting the demands of an ever increasing number of requests, the easiest solution is to integrate more advanced hardware to existing systems. However, additional investment in hardware may significantly increase the system cost as well as daily power consumption. In this paper, we present evidence that well-selected software level optimization is capable of achieving comparable levels of performance without the cost and power consumption overhead caused by physically expanding the system. Specifically, we develop intelligent prefetching algorithms that are suitable for the unique workloads and user behaviors of the world's largest satellite images distribution system managed by USGS EROS. Our experimental results, derived from real-world traces with over five million requests sent by users around the globe, show that the EROS hybrid storage system could maintain the same performance with over 30% of energy savings by utilizing our proposed prefetching algorithms, compared to the alternative solution of doubling the size of the current FTP server farm.

Featured by high portability and programmability, Dynamic Voltage and Frequency Scaling (DVFS) has been widely employed to achieve energy efficiency for high performance applications on distributed-memory architectures nowadays through various scheduling algorithms. Generally, different forms of slack from load imbalance, network latency, communication delay, memory and disk access stalls, etc. are exploited as energy saving opportunities where peak CPU performance is not necessary, with little or limited performance loss. The deployment of DVFS for communication intensive applications is straightforward due to the explicit boundary between Energy Saving Blocks (ESBs) at source code level, while for data (e.g., memory and disk access) intensive applications it is difficult for applying DVFS since ESB boundary is implicit due to mixed types of workloads. We propose an adaptively aggressive DVFS scheduling strategy to achieve energy efficiency for data intensive applications, and further save energy via speculation to mitigate DVFS overhead for imbalanced branches. We implemented and evaluated our approach using five memory and disk access intensive benchmarks with imbalanced branches against another two energy saving approaches. The experimental results indicate an average of 32.6% energy savings were achieved with 6.2% average performance loss compared to the original executions on a power-aware 64-core cluster.

N-body simulations are computation-intensive applications that calculate the motion of a large number of bodies under pair-wise forces. Although different versions of n-body codes have been widely used in many scientific fields, the performance and energy efficiency of various n-body codes have not been comprehensively studied, especially when they are running on newly released multi-core CPUs and GPUs (e.g., Tesla K20). In this paper, we evaluate the performance and energy efficiency of five parallel n-body implementations on two different multi-core CPU systems and on two different types of GPUs. Our experimental results show that up to 71% of the energy can be saved by using all cores of a Xeon E5620 CPU instead of only one. We find hyper-threading to be able to further reduce the energy usage and runtime, but not by as much as adding more cores does. Finally, our experiments illustrate that GPU-based acceleration using a Tesla K20c can boost the performance and energy efficiency by orders of magnitude.

Online content distribution systems, which store incredibly large amounts of information and provide service to large numbers of users, are becoming increasingly commonplace. To fulfill the wide range of requests sent by different users, these systems must ensure efficient handling of massive amount of data. To achieve this goal, the in-depth analysis and comprehensive understanding of user behaviors are critical. However, analyzing the behaviors of worldwide users with different needs is a very challenging task. This is especially true when historical user behaviors evolve over time or may be affected by unpredictable events. In this paper, we present a number of workload characterization techniques applied to one of the world's largest online satellite image distribution systems operated by the U.S. Geological Survey (USGS) and NASA.

Most researches of Solid State Drives (SSDs) architectures rely on Flash Translation Layer (FTL) algorithms and wear-leveling; however, internal parallelism in Solid State Drives has not been well explored. In this research, we proposed a new strategy to improve SSD write performance by enhancing internal parallelism inside SSDs. A SDRAM buffer is added in the design for buffering and scheduling write requests. Because the same logical block numbers may be translated to different physical numbers at different times in FTL, the on-board SDRAM buffer is used to buffer requests at the lower level of FTL. When the buffer is full, same amount of data will be assigned to each storage package in SSDs to enhance internal parallelism. To accurately evaluate performance, we use both synthetic workloads and real-world applications in experiments. We compare the enhanced internal parallelism scheme with the traditional LRU strategy since it is unfair to compare an SSD having buffer with an SSD without a buffer. The simulation results demonstrate that the writing performance of our design is significantly improved compared with the LRU-cache strategy with the same amount of buffer sizes.

Caching is widely known to be an effective method for improving I/O performance by storing frequently used data on higher speed storage components. However, most existing studies that focus on caching performance evaluate fairly small files populating a relatively small cache. Few reports are available that detail the performance of traditional cache replacement policies on extremely large caches. Do such traditional caching policies still work effectively when applied to systems with petabytes of data? In this paper, we comprehensively evaluate the performance of several cache policies, which include First-In-First-Out (FIFO), Least Recently Used (LRU) and Least Frequently Used (LFU), on the global satellite imagery distribution application maintained by the U.S. Geological Survey (USGS) Earth Resources Observation and Science Center (EROS). Evidence is presented suggesting traditional caching policies are capable of providing performance gains when applied to large data sets as with smaller data sets. Our evaluation is based on approximately three million real-world satellite images download requests representing global user download behavior since October 2008.

The United States is currently pursuing renewable energy research and education initiatives. To contribute to both of these objectives, an interactive environment that provides educational opportunities related to wind energy for students in K-12, college, and the community was created. The efforts are based on an 80 ft tall 20kW wind turbine that was installed at the South Dakota School of Mines and Technology (SDSMT) in 2009. This research facility is currently recording real-time wind data that is not easily accessed or understood by the public community due to the absence of an ease-of-communication environment. To increase the educational outreach of the wind turbine facility, an interactive computer kiosk in the student center at SDSMT was designed and utilized. The functionality of the kiosk includes: 1) a 24/7 searchable real-time database; 2) a 3D visual model of the wind turbine and research facility; and 3) a user-created energy consumption “sandbox”. The results of the study will be used to provide recommendations for future research and education endeavors of renewable energy in general and wind-energy in particular.

Mobile technologies are ubiquitous in modern society, and college students find themselves increasingly immersed in their use. Is it possible to integrate this hand-held universe of mobile technology into the STEM classroom so that it becomes an organic, integral part of the 21st-century learning experience? Can mobile technology be used to enhance and improve the learning experience of students, one that develops both their technical and psycho-social skills? This paper discusses our attempt to do so at the South Dakota School of Mines and Technology, by developing a “mobile portal” through which instructors and students can access mobile content to supplement and expand the in-class learning environment.

Performance improvement and energy conservation are two conflicting objectives in large scale parallel storage systems. In this paper, we propose a novel solution to achieve the twin objectives of maximizing performance and minimizing energy consumption of parallel storage systems. Specifically, a buffer-disk based architecture (BUD for short) is designed to conserve energy. A heat-based dynamic data caching strategy is developed to improve performance. The BUD architecture strives to allocate as many requests as possible to buffer disks, thereby keeping a large number of idle data disks in low-power states. This can provide significant opportunities for energy conservation while making buffer disks a potential performance bottleneck. The heat-based data caching strategy aims to achieve good load balancing in buffer disks and alleviate overall performance degradation caused by unbalanced workload. Our experimental results have shown that the proposed BUD framework and dynamic data caching strategy are able to conserve energy by 84.4% for small reads and 78.8% for large reads with slightly degraded response time.

High-performance clusters have been widely deployed to solve challenging and rigorous scientific and engineering tasks. On one hand, high performance is certainly an important consideration in designing clusters to run parallel applications. On the other hand, the ever increasing energy cost requires us to effectively conserve energy in clusters. To achieve the goal of optimizing both performance and energy efficiency in clusters, in this paper, we propose two energy-efficient duplication-based scheduling algorithms-Energy-Aware Duplication (EAD) scheduling and Performance-Energy Balanced Duplication (PEBD) scheduling. Existing duplication-based scheduling algorithms replicate all possible tasks to shorten schedule length without reducing energy consumption caused by duplication. Our algorithms, in contrast, strive to balance schedule lengths and energy savings by judiciously replicating predecessors of a task if the duplication can aid in performance without degrading energy efficiency. To illustrate the effectiveness of EAD and PEBD, we compare them with a nonduplication algorithm, a traditional duplication-based algorithm, and the dynamic voltage scaling (DVS) algorithm. Extensive experimental results using both synthetic benchmarks and real-world applications demonstrate that our algorithms can effectively save energy with marginal performance degradation.

Improving security and minimizing power consumption are crucial for large-scale data storage systems. Although a handful of studies have been focused on data security and energy efficiency, most of the existing approaches have concentrated on only one of these two metrics. In this paper, we present a new approach to integrating power optimization with security services to enhance the security of energy-efficient large-scale storage systems. In our approach, we make use of the dynamic speed control for power management technique, or DRPM, to conserve energy in secure storage systems. In this study we develop two ways of integrating confidentiality services with the dynamic disk speed control technique. The first strategy - security aggressive in nature - is focused on the improvement of storage system security with less emphasis on energy conservation. The second strategy gives higher priority to energy conservation as opposed to the security optimization. Our experimental results show that the energy-aggressive approach provides better energy savings than the security-aggressive approach. However, the quality of security achieved by the security-aggressive scheme is higher than that of the energy-aggressive approach. Moreover, the empirical results show that energy savings yielded by the two approaches become more pronounced when the data size is increased. The findings illustrate that the response time of the security-aggressive approach is more sensitive to data size than that of the energy-aggressive scheme.

An infrastructure-as-a-service cloud system provides computational capacities to remote users. Parallel processing in the cloud system can shorten the execution of jobs. Parallel processing requires a mechanism to scheduling the executions order as well as resource allocation. Furthermore, a preemptable scheduling mechanism can improve the utilization of resources in clouds. In this paper, we present a preemptable job scheduling mechanism in cloud system. We propose two feedback dynamic scheduling algorithms for this scheduling mechanism. We compare these two scheduling algorithms in simulations. The results show that the feedback procedure in our algorithms works well in the situation where resource contentions are fierce.

In the past decade, parallel disk systems have been developed to address the problem of I/O performance. A critical challenge with modern parallel I/O systems is that parallel disks consume a significant amount of energy in servers and high performance computers. To conserve energy consumption in parallel I/O systems, one can immediately spin down disks when disk are idle; however, spinning down disks might not be able to produce energy savings due to penalties of spinning operations. Unlike powering up CPUs, spinning down and up disks need physical movements. Therefore, energy savings provided by spinning down operations must offset energy penalties of the disk spinning operations. To substantially reduce the penalties incurred by disk spinning operations, we developed a novel approach to conserving energy of parallel I/O systems with write buffer disks, which are used to accumulate small writes using a log file system. Data sets buffered in the log file system can be transferred to target data disks in a batch way. Thus, buffer disks aim to serve a majority of incoming write requests, attempting to reduce the large number of disk spinning operations by keeping data disks in standby for long period times. Interestingly, the write buffer disks not only can achieve high energy efficiency in parallel I/O systems, but also can shorten response times of write requests. To evaluate the performance and energy efficiency of our parallel I/O systems with buffer disks, we implemented a prototype using a cluster storage system as a testbed. Experimental results show that under light and moderate I/O load, buffer disks can be employed to significantly reduce energy dissipation in parallel I/O systems without adverse impacts on I/O performance.

Modern real-time wireless networks require high security level to assure confidentiality of information stored in packages delivered through wireless links. However, most existing algorithms for scheduling independent packets in real-time wireless networks ignore various security requirements of the packets. Therefore, in this paper we remedy this problem by proposing a novel dynamic security-aware packet-scheduling algorithm, which is capable of achieving high quality of security for realtime packets while making the best effort to guarantee realtime requirements (e.g., deadlines) of those packets. We conduct extensive simulation experiments to evaluate the performance of our algorithm. Experimental results show that compared with two baseline algorithms, the proposed algorithm can substantially improve both quality of security and real-time packet guarantee ratio under a wide range of workload characteristics.

Real time applications such as military aircraft flight control systems and online banking are critical with respect to security and reliability. In this paper we presented a way to integrate both by considering confidentiality and integrity services for security and nonuniform checkpoint strategy for reliability. The slack exploitation interacts in subtle ways for security in regards to the placement of checkpoint. The checkpoints are placed in to the task at low frequency in the beginning because the slack available can accommodate a large amount of work at risk and the frequency is increased there after considering the slack available. The security is applied to the data in two ways. First method introduces the security for the entire data at once whereas in the second method the data is divided into n uneven sections and each section is separately secured . That is at the start of the task basic security services are considered depending on the slack available. The security is increased gradually for the rest of the task but if there exist a fault, then at that point the security is maintained at the steady rate because of the limited slack. Compared to the first method the second method can provide up to a 32.3 percent higher security. While compared to the traditional checkpoint strategy, the non-uniform check pointing makes more efficient use of slack while increasing the overall security levels by 34.4 percent for the second method.

In the past decade cluster computing platforms have been widely applied to support a variety of scientific and commercial applications, many of which are parallel in nature. However, scheduling parallel applications on large scale clusters is technically challenging due to significant communication latencies and high energy consumption. As such, shortening schedule length and conserving energy consumption are two major concerns in designing economical and environmentally friendly clusters. In this paper, we propose an energy-efficient scheduling algorithm (TDVAS) using the dynamic voltage scaling technique to provide significant energy savings for clusters. The TDVAS algorithm aims at judiciously leveraging processor idle times to lower processor voltages (i.e., the dynamic voltage scaling technique or DVS), thereby reducing energy consumption experienced by parallel applications running on clusters. Reducing processor voltages, however, can inevitably lead to increased execution times of parallel task. The salient feature of the TDVAS algorithm is to tackle this problem by exploiting tasks precedence constraints. Thus, TDVAS applies the DVS technique to parallel tasks followed by idle processor times to conserve energy consumption without increasing schedule lengths of parallel applications. Experimental results clearly show that the TDVAS algorithm is conducive to reducing energy dissipation in large-scale clusters without adversely affecting system performance.

High performance clusters have been widely used to provide amazing computing capability for both commercial and scientific applications. However, huge power consumption has prevented the further application of large-scale clusters. Designing energy-efficient scheduling algorithms for parallel applications running on clusters, especially on the high performance heterogeneous clusters, is highly desirable. In this regard, we propose a novel scheduling strategy called energy efficient task duplication schedule (EETDS for short), which can significantly conserve power by judiciously shrinking communication energy cost when allocating parallel tasks to heterogeneous computing nodes. We present the preliminary simulation results for Gaussian and FFT parallel task models to prove the efficiency of our algorithm.

Existing packets scheduling algorithms designed for energy-efficient wireless networks ignore important features of periodic packets, thereby being inadequate for periodic packets with energy constraints. To remedy this problem, we present in this paper an approach to scheduling periodic packets in wireless networks subject to both timing and energy constraints. We propose a necessary and sufficient feasibility check for a set of periodic packets to be transmitted over a wireless link. Next, we develop an algorithm to schedule periodic packets (or ESPP for short) over a wireless link. The ESPP algorithm aims at minimizing energy dissipation of periodic packets without missing deadlines of periodic packets. We show through simulation studies that ESPP can significantly reduce energy consumption of wireless networks by an average of 46.4% while guaranteeing timing constraints of periodic packets.

Huge energy consumption has become a critical bottleneck for further applying large-scale cluster systems to build new data centers. Among various components of a data center, storage subsystems are one of the biggest consumers of energy. In this paper, we propose a novel buffer-disk based framework for large-scale and energy-efficient parallel storage systems. To validate the efficiency of the proposed framework, a buffer-disk scheduling algorithm is designed and implemented. Our algorithm can provide more opportunities for underlying disk power management schemes to save energy by keeping a large number of idle data disks in sleeping mode as long as possible. The trace-driven simulation results based on a revised disksim simulator show that this new framework can significantly improves the energy efficiency of large-scale parallel storage systems.

Optimizing energy consumption has become a major concern in designing economical clusters. Scheduling precedence-constrained parallel tasks on clusters is challenging because of high communication overhead. Although duplication-based strategies are applied to minimize communication overhead, most of them merely consider schedule lengths, completely ignoring energy consumption of clusters. In this regard, we propose two energy-aware duplication scheduling algorithms, called EADUS and TEBUS, to schedule precedence-constrained parallel tasks. Unlike existing duplication-based scheduling algorithms that replicate all possible predecessors of each task, the proposed algorithms judiciously replicate predecessors only if the duplication can help in conserving energy. Our energy-aware scheduling strategies are conducive to balancing the scheduling length and energy consumption of precedence-constrained parallel tasks. Extensive experimental results based on real-world applications demonstrate the effectiveness and practicality of the proposed scheduling strategies