Poster部分-2016年全国计算机体系结构学术年会

 ACA2016ContributedTalk,Poster&ExhibitionSession摘要集（Poster部分）8月23日 1. Micro-architectural Features for Malware Detection Huicheng Peng, Jizeng Wei and Wei Guo Tianjin Advanced Network Key Lab Abstract: As the variety and complexity of attacks continue to increase, software-based malware detection can impose significant performance overhead. Recent works have demonstrated the feasibility of malware detection using hardware performance counters. Therefore, equipping a malware detector to collect and analyze micro-architecture features of CPUs to recognize malware at running time has become a promising method. In comparison to the software-based malware detection, hardware-based malware detection not only reduces the cost of system performance, but also possesses better detection capacity. However, hundreds of micro-architecture events can be monitored by hardware performance counters (HPCs) which are widely available in prevailing CPUs, such as Intel, ARM and so on. In this paper, we take Intel ivy bridge i3 processor as an example and examine most of these micro-architectural features. Instead of relying on experience, the Lasso algorithm is employed to reduce the dimensionality of feature vector to 6 elements. Furthermore, 4 classification methods based on supervised learning are applied for the selected features. We improve the classification accuracy rate of 15% on average. The results show that the micro-architectural features of this paper can reveal the behaviors of malware better. 2. H-TDMS: A System for Traffic Big Data Management Xingcheng Hua, Jierui Wang, Li Lei, Bin Zhou, Xiaolin Zhang and Peng Liu Zhejiang University Abstract: Massive traffic data is produced constantly every day, causing problems in data integration, massive storage, high performance processing when applying conventional data management approaches. We propose a cloud computing based system H-TDMS (Hadoop based Traffic Data Management System) to capture, manage and process the traffic big data. H-TDMS designs a configurable tool for data integration, a scalable data scheme for data storage, a secondary index for fast search query, a computing framework for data analysis, and a web-based user-interface with data visualization service for user interaction. Experiments on actual traffic data show that H-TDMS achieves considerable performance in traffic big data management. 3. 集成IO硬件压缩加速器的Hadoop系统结构 雷力，钱斌海，郭俊，顾雄礼，刘鹏 浙江大学 摘要：随着大数据的发展，Hadoop系统成为了大数据处理中的重要工具之一。在实际应用中，Hadoop的I/O操作制约系统性能的提升。通常Hadoop系统通过软件压缩数据来减少I/O操作，但是软件压缩速度较慢，因此使用硬件压缩加速器来替换软件压缩。Hadoop运行在Java虚拟机上，无法直接调用底层I/O硬件压缩加速器。通过实现Hadoop压缩器/解压缩器类和设计C++动态链接库来解决从Hadoop系统中获得压缩数据和将数据流向I/O硬件压缩加速器两个关键技术，从而将I/O硬件压缩加速器集成到 Hadoop系统框架。实验结果表明，I/O硬件压缩加速器的每赫兹压缩速度为15.9Byte/s/Hz，集成I/O硬件压缩加速器提升Hadoop系统性能2倍。 4. GLDA: Parallel Gibbs Sampling for Latent Dirichlet Allocation on GPU Pei Xue, Tao Li, Kezhao Zhao, Qiankun Dong, Wenjing Ma Nankai University Abstract: With the development of the general computing ability of GPU, more and more algorithms are being run on GPU, to enjoy much higher speed. In this paper, we propose an approach that uniformly accelerate Gibbs sampling for LDA (Latent Dirichlet Allocation) Algorithm on GPU, which makes the data load to the cores of GPU evenly to avoid the idle waiting for GPU, and improves the utilization of GPU. We use three text mining datasets to test the algorithm. Experiments show that our parallel methods can achieve about 30x speedup over sequential training methods with similar prediction precision. Furthermore, the idea that uniformly partitioning the data bases on GPU can also be applied to other machine learning algorithms. 5. OPTAS: Decentralized Flow Monitoring and Scheduling for Tiny Tasks Ziyang Li, Yiming Zhang, Dongsheng Li and Yuxing Peng National University of Defense Technology Abstract: Task-aware flow schedulers collect task information across the data center to optimize task-level performance. However, the majority of the tasks, which generate short flows and are called tiny tasks, have been largely overlooked by current schedulers. The large number of tiny tasks brings significant overhead to the centralized schedulers, while the existing decentralized schedulers are too complex to fit in commodity switches.In this paper we present OPTAS, a lightweight, commodity-switch-compatible scheduling solution that efficiently monitors and schedules flows for tiny tasks with low overhead. OPTAS monitors system calls and buffer footprints to recognize the tiny tasks, and assigns them with higher priorities than larger ones. The tiny tasks are then transferred in a FIFO manner by adjusting two attributes, namely, the window size and round trip time, of TCP. We have implemented OPTAS as a Linux kernel module, and experiments on our 37-server testbed show that OPTAS is at least 2.2× faster than fair sharing, and 1.2× faster than only assigning tiny tasks with the highest priority. 6. A Model for Evaluating and Comparing Movin