Tseng-Yi Chen

LaSA: A locality-aware scheduling algorithm for Hadoop-MapReduce resource assignment

Cloud computing has become more popular for a decade; it has been under continuous development with advances in architecture, software, and network. Hadoop-MapReduce is a common software framework processing parallelizable problem across big datasets using a distributed cluster of processors or stand-alone computers. Cloud Hadoop-MapReduce can scale incrementally in the number of processing nodes. Hence, the Hadoop-MapReduce is designed to provide a processing platform with powerful computation. Network traffic is always a most important bottleneck in data-intensive computing and network latency decreases significant performance in data parallel systems. Network bottleneck is caused by network bandwidth and the network speed is much slower than disk data access. So that, good data locality can reduces network traffic and increases performance in data-intensive HPC systems. However, Hadoops scheduler has a defect of data locality in resource assignment. In this paper, we present a locality-aware scheduling algorithm (LaSA) for Hadoop-MapReduce scheduler. Firstly, we propose a mathematical model of weight of data interference in Hadoop scheduler. Secondly, we present the LaSA algorithm to use weight of data interference to provide data locality-aware resource assignment in Hadoop scheduler. Finally, we build an experimental environment with 3 cluster and 35 VMs to verify the LaSAs performance.

CacheRAID: An Efficient Adaptive Write Cache Policy to Conserve RAID Disk Array Energy

Cloud storage is a hot topic at the moment with Googles Google Storage, Microsofts Sky Drive, iCloud, Drop box, Mozy and dozens of others. Because of these applications, conserving energy of storage systems is becoming a growing concern in current storage technology. The factors of disk power consumption include disk idle time, poor random writing performance and random read in distributed file systems. Hence, we present an adaptive write cache mechanism - Cache RAID. Redundant Arrays of Inexpensive Disk (RAID) is widely used in modern distributed storage systems. Our Cache RAID aims to improve the random access problems that implicitly exist in RAID techniques to create more idle time of hard drives, and conserve RAID disk array energy. The experimental results show that Cache RAID storage system can conserved 50%~70% of the power consumption compared to the conventional software RAID system.

Efficient Parallel Algorithm for Nonlinear Dimensionality Reduction on GPU

Advances in nonlinear dimensionality reduction provide a way to understand and visualize the underlying structure of complex data sets. The performance of large-scale nonlinear dimensionality reduction is of key importance in data mining, machine learning, and data analysis. In this paper, we concentrate on improving the performance of nonlinear dimensionality reduction using large-scale data sets on the GPU. In particular, we focus on solving problems including k nearest neighbor (KNN) search and sparse spectral decomposition for large-scale data, and propose an efficient framework for Local Linear Embedding (LLE). We implement a k-d tree based KNN algorithm and Krylov subspace method on the GPU to accelerate the nonlinear dimensionality reduction for large-scale data. Our results enable GPU-based k-d tree LLE processes of up to about 30-60 X faster compared to the brute force KNN [10] LLE model on the CPU. Overall, our methods save O (n2-6n-2k-3) memory space.

Enabling Write-Reduction Strategy for Journaling File Systems over Byte-addressable NVRAM

Non-volatile random-access memory (NVRAM) becomes a mainstream storage device in embedded systems due to its favorable features, such as small size, low power consumption, and short read/write latency. On NVRAM, a write operation consumes more energy and time than a read operation. However, current mobile/embedded file systems (e.g., EXT2/3 and EXT4) are very unfriendly for NVRAM devices. The reason is that a journaling mechanism writes the same data twice during data commitment and checkpoint. Such observations motivate this paper to design a two-phase write reduction journaling file system called wrJFS. In the first phase, wrJFS classified data into two categories: Metadata and user data. Metadata will be handled by partial byte-enabled journaling strategy, and user data will be processed in the second phase. In the second phase, user data will be compressed by hardware encoder so as to reduce the write size, and managed compressed-enabled journaling strategy to avoid the write amplification. The experimental results show that the proposed wrJFS can reduce the size of the write request by 89.7% on average, compared with the original EXT3.

An effective monitoring framework and user interface design

A distributed environment requires a monitoring system to oversee the operation of various distributed nodes. A monitoring service is crucial because it ensures a high‐quality computing environment and a reliable service. The interface and framework determine the effectiveness of a monitoring system. This paper uses the concept of user‐adaptive visualization to design its interface and proposes a flexible modular framework. Designers can use the proposed modular framework to flexibly extend existing modules, design visual interfaces to satisfy user requirements, and improve system failover schemes. The implementation of such a monitoring system for monitoring data preservation nodes is also provided. The system including fault‐tolerance and notification functions supports full monitoring services for Storage Resource Broker (SRB) or integrated Rule‐Oriented Data System (iRODS) based systems. The experimental results show that the proposed framework is suitable for data preservation services and is robust and responsive when faced with system failures. Copyright

Optimizing space utilization of file systems on PCM-based storage devices

Phase-change memory (PCM) is a promising candidate as a storage medium to resolve the performance gap between main memory and storage in battery-powered mobile computing systems. However, it is more expensive than flash memory, and thus introduces a more serious storage capacity issue in low-cost solutions. This issue is further exacerbated by the fact that existing file systems are usually designed to trade space utilization for performance over block-oriented storage devices. In this work, we propose a multi-grained block management scheme to optimize the space utilization of file systems on PCM-based storage systems. By utilizing the byte-addressability and fast read/write feature of PCM, a methodology is proposed to dynamically allocate multiple sizes of blocks to fit the size of each file, so as to resolve the space fragmentation issue with minimized space and management overheads. A series of experiments was conducted to evaluate the efficacy of the proposed scheme, and the results show that the proposed scheme could significantly improve the space utilization of file systems.

EEGRA: Energy Efficient Geographic Routing Algorithms for Wireless Sensor Network

Energy efficiency is critical in wireless sensor networks (WSN) for system reliability and deployment cost. The power consumption of the communication in multi-hop WSN is primarily decided by three factors: routing distance, signal interference, and computation cost of routing. Several routing algorithms designed for energy efficiency or interference avoidance had been proposed. However, they are either too complex to be useful in practices or specialized for certain WSN architectures. In this paper, we propose two energy efficient geographic routing algorithms (EEGRA) for wireless sensor networks, which are based on existing geographic routing algorithms and take all three factors into account. The first algorithm combines the interference into the routing cost function, and uses it in the routing decision. The second algorithm transforms the problem into a constrained optimization problem, and solves it by searching the optimal discretized interference level. We integrate four geographic routing algorithms: GOAFR+, Face Routing, GPSR, and RandHT, to both EEGRA algorithms and compare them with three other routing methods in terms of power consumption and computation cost for the grid and irregular sensor topologies. The results of our experiments show both algorithms conserve sensors routing energy 30% ~ 50% comparing to general geographic routing algorithms. In addition, the time complexity of EEGRA algorithms is similar to the geographic greedy routing methods, which is much faster than the optimal SINR-based algorithm.

Enabling sub-blocks erase management to boost the performance of 3D NAND flash memory

3D NAND has been proposed to provide a large capacity storage with low-cost consideration due to its high density memory architecture. However, 3D NAND needs to consume enormous time for garbage collection because of live-page copying overhead and long block erase time. To alleviate the impact of live-page copying on the performance of 3D NAND, a sub-block erase design has been designed. With sub-block erase design, this paper proposes a performance booster strategy to extremely boost the performance of garbage collection. As experimental results shows, the proposed strategy has a significant improvement on the average response time.

An Energy-Efficient and Reliable Storage Mechanism for Data-Intensive Academic Archive Systems

Previous studies proposed energy-efficient solutions, such as multispeed disks and disk spin-down methods, to conserve power in their respective storage systems. However, in most cases, the authors did not analyze the reliability of their solutions. According to research conducted by Google and the IDEMA standard, frequently setting the disk status to standby mode will increase the disk’s Annual Failure Rate and reduce its lifespan. To resolve the issue, we propose an evaluation function called E3SaRC (Economic Evaluation of Energy Saving with Reliability Constraint), which considers the cost of hardware failure when applying energy-saving schemes. We also present an adaptive write cache mechanism called CacheRAID. The mechanism tries to mitigate the random access problems that implicitly exist in RAID techniques and thereby reduce the energy consumption of RAID disks. CacheRAID also addresses the issue of system reliability by applying a control mechanism to the spin-down algorithm. Our experimental results show that the CacheRAID storage system can reduce the power consumption of the conventional software RAID 5 system by 65% to 80%. Moreover, according to the E3SaRC measurement, the overall saved cost of CacheRAID is the largest among the systems that we compared.

BASE: Benchmark analysis software for energy-efficient solutions in large-scale storage systems

The concept of green storage in cluster computing has generated a great deal of interest among researchers in recent years. As a result, several energy-efficient solutions, such as multi-speed disks and disk spin down methods, have been proposed to conserve power in storage systems and improve disk access. Some researchers evaluate their solutions via simulations, while others utilize real-world experiments. Both methods have advantages and disadvantages. To address the problem, we propose an efficient simulation tool called BASE, which can accurately estimate the power consumption of disks in large-scale storage systems. We evaluate the performance of BASE on real-world traces from Academia Sinica (Taiwan) and Florida International University. BASE incorporates an analytical method for evaluating the reliability of energy-efficient solutions. Our analysis results show that the measurement error of BASE is 2.5% lower than that achieved in real-world experiments on energy estimation. Moreover, the results of simulations performed to evaluate a solutions reliability are the same as those derived by real-world experiments.