基于软件定义存储的小文件性能优化研究

针对现有技术中对小文件性能优化的效率低、运算不稳定等问题,研究出适用的解决方案。该方案基于OpenStack云平台构建软件定义对象存储系统,其中,分类合并处理模块采用分类合并方法和改进的GDS缓存替换算法,以实现缓存空间的合理分配与使用。动态预处理模块采用What-if假设方法并融合关联规则技术进行改进,提高对多种结构的小文件数据处理效率。试验表明,文中的存储系统中小文件的平均读写响应时间分别减少了32.3%和25.1%,适用性和可靠性高。     

针对Flash存储特性的航天器大容量固态存储技术

航天任务对存储系统要求极高,商用领域基于Flash的固态硬盘使用FTL技术照搬传统磁盘管理技术,对Flash存储特性利用不足,无法满足需求.本文基于3Dplus公司的Flash存储模块,采用FPGA直接管理存储介质的方法,结合航天任务需求,针对Flash存储特性提出并行总线、流水线操作技术和坏块管理等航天应用措施,介绍一种实际应用于航天任务的大容量固态存储系统,容量128 Gbits,吞吐率可达500 Mbps.最后讨论了民用Flash存储系统的发展对航天应用的启示.     

云存储环境下的混合存储算法研究与实现

随着大数据的发展,数据运算、数据读写均对存储环境和存储系统提出了更高的要求。存储系统中所运行的任务具有高并发度特征,也就是要对其存储的数据实现随机读写的性能。固态盘(简称SSD)是一种云储存环境下的新型存储设备,具有随机读写数据的功能,但是价格比较昂贵,而且对于写入的次数有限制,所以应该把固态盘存储技术和磁盘存储(HDD)结合起来,形成云存储环境下的混合存储系统。文章在分析三种基本存储设备特点的基础上,重点阐述在云存储环境下,混合存储算法的研究与实现。     

固态硬盘作为制播网存储系统中数据存储设备探讨

在非编网和播出网的存储系统中,传统的机械硬盘已经严重影响了存储系统的性能.本文就目前流行的固态硬盘作为存储系统的存储设备提出了一些看法.     

面向大数据的分布式缓存设计

大数据平台的底层存储系统往往无法匹配上层计算应用的读写性能,而一个设计良好的分布式缓存系统将缩小CPU密集型应用和IO密集型应用之间不匹配的性能差距。设计的面向大数据应用的分布式缓存系统,在读写流程、I/O事件驱动并发模型及元数据模型等方面进行了合理设计与优化,并使用fio工具测试了顺序写、随机写、顺序读及随机读场景下的吞吐率与IOPS等性能指标,验证了该分布式缓存系统的高性能优势和应对高并发场景的扩展能力。     

面向异构并行计算系统的流水线式压缩检查点

在大规模并行计算系统中,并行检查点触发大量结点同时保存计算状态,造成巨大文件存储空间开销,以及对通信和存储系统的巨大访问压力.数据压缩可以缩小检查点文件尺寸,从而降低存储空间开销以及对通信和存储系统的访问压力.但是,它也带来额外的压缩计算开销.本文针对异构并行计算系统,提出流水线式并行压缩检查点技术,采用一系列优化技术来降低压缩引入的计算延时,包括:流水线式双重写缓存队列、文件写操作的合并、GPU加速的流水压缩算法和GPU资源的多进程调度,等等.本文介绍了该技术在天河一号系统中的实现,并对所实现的检查点系统进行综合评测.实验数据表明该方法在大规模异构并行计算系统中是可行、高效、实用的.     

基于闪存阵列的缓存容量确定方法

针对低成本、小型化的数据记录系统的应用,提出了一种数据缓存技术解决方案。存储模块是由闪速存储器芯片（NAND Flash）组成的存储阵列,以FPGA为载体的SOPC系统作为存储模块的控制核心。分析存储系统的结构及控制平台的实现过程,并对系统工作原理及并行分路技术进行讨论。深入研究Flash阵列的存储过程,提出最小FIF...     

一种超大容量自动光盘库的设计与实现

当前蓝光光盘的寿命已超过50年,光存储的可靠性远高于硬盘,寿命也远长于磁带,但单盘容量较小、存取性能较低的缺点限制了光盘在大规模归档系统中的应用。提出了一种新型超大容量机械手自动换盘的光盘库系统,该系统能够在标准尺寸的机柜中容纳12 000张蓝光光盘,数十个光驱可并行读写,对外的吞吐率达到1 GB/s。除了高度并行之外,还使用了磁光电融合结构和虚拟化存储机制,通过磁电作为光存储的大容量缓存,提高存取性能,将大量的光盘存储空间虚拟成单个文件卷存储池。该系统的光盘调度、刻录和读取完全实现自动化,并提供给用户通用文件访问接口。综合这些技术,既发挥了光存储介质的大容量、长寿命、低成本、低能耗的优点,又克服了光存储系统速度慢、性能低的缺点,同时提供了用户友好的使用界面和环境,实现了与现有信息系统的无缝对接。     

云存储系统中基于更新日志的元数据缓存同步策略

在云存储系统中使用客户端元数据缓存,不仅可以减轻元数据服务器的压力,避免元数据服务器成为系统瓶颈,提高系统的可扩展性,而且可以降低客户端元数据操作的响应延迟,极大地提高客户端IOPS性能.本文对云存储系统客户端元数据缓存同步机制进行研究,提出了基于更新日志的元数据缓存同步策略,利用该策略替代了MassCloud云存储系统原先的元数据缓存完全同步策略.测试结果表明,新的缓存同步策略给MassCloud云存储系统带来了20～140倍的性能提升.     

面向类仿射型数组下标应用的参数化并行存储结构模板

为了解决目前可重构编译技术在为类仿射型数组下标应用生成循环流水阵列时,生成的存储系统对数据并行与重用支持不完善的问题,本文提出了一种参数化并行存储结构模板。此模板采用模块化设计思想,根据数据访存特征生成由多体交叉并行存储子模块、单体串行存储子模块、RAW Buffer缓存子模块及Smart Buffer缓存子模块构成的存储结构。为灵活生成存储结构及充分挖掘数据的并行性和重用性,本文采用访存数据依赖图方法计算存储模板的参数值。和相关工作相比,根据本文提出的存储结构模板生成的硬件,可以在占用较少的硬件资源情况下,获得较高的硬件执行速度。     

状态空间法在高分辨雷达中的应用研究

高分辨雷达一般采用两套系统，窄带系统作为传统的坐标测量系统，实现目标检测和跟踪；宽带系统实现成像和识别，其在窄、宽带系统之间转换的工作方式会降低雷达的数据率。该文将状态空间方法引入基于宽带一维高分辨距离像的单脉冲雷达的坐标测量，从而实现了宽带系统中的测距、测角，在抑制角闪烁的同时，提高了扩展目标上各散射中心的坐标测量精度。     

Scalability and Security Conflict for RFID Authentication Protocols

RFID technology continues to flourish as an inherent part of virtually every ubiquitous environment. However, it became clear that the public—implying the industry—seriously needs mechanisms emerging the security and privacy issues for increasing RFID applications. As the nodes of RFID systems mostly suffer from low computational power and small memory size, various attempts which propose to implement the existing security primitives and protocols, have ignored the realm of the cost limitations and failed. In this study, two recently proposed protocols—SSM and LRMAP—claiming to meet the standard privacy and security requirements are analyzed. The design of both protocols based on defining states where the server authenticates the tag in constant time in a more frequent normal state and needs a linear search in a rare abnormal states. Although both protocols claim to provide untraceability criteria in their design objectives, we outline a generic attack that both protocols failed to fulfill this claim. Moreover, we showed that the SSM protocol is vulnerable to a desynchronization attack which prevents a server from authenticating a legitimate tag. Resultantly, we conclude that defining computationally unbalanced tag states yields to a security/scalability conflict for RFID authentication protocols.     

Linearity and CNR Improvement Technologies against Optical Reflection in Fiber-Oriented Wireless Access Systems

It is important to consider optical reflectionconditions when designing fiber-oriented wirelessaccess systems around subcarrier optical transmissionsince the noise characteristic can be significantlydegraded by optical reflection, especially if narrowspectral width devices are used. The superluminescentdiode (SLD) offers stable noise and 3rd-orderintermodulation distortion (IM3) characteristics thatfollow the ideal cubic law, even under multipleoptical reflection. The SLD-predistorter combinationis proposed as a way to overcome the SLD's narrowdynamic range. It is experimentally found that thepredistorter reduces IM3 by 8 dB. The receiversensitivity of a practical system that uses theSLD-predistorter combination is discussed from theviewpoint of overall design. A superimposedsubcarrier modulation (SSM) technique is also proposedin order to reduce the reflection noise. Weexperimentally confirm that the SSM technique reducesthe noise degradation and provides stable opticaltransmission. An SSM design methodology is presentedthat takes both noise and linearity into account. TheSLD-predistorter and the SSM scheme make it possibleto realize stable transmission performance and widerdynamic range in the presence of optical reflection.     

Simulation of error sources in digital channels

A general approach to error-source simulation in digital channels is discussed. It is shown that the stochastic sequential machine (SSM) model is general enough to describe error-source statistics in digital communication channels. The SSM model is capable of describing both the correlation between errors in a channel (bursts of errors) and the correlation between errors in different channels (crosstalk). Physical causes of error burstiness in satellite channels are considered, and it is shown that the SSM model can account for differential encoding, scrambling, forward error correction, intersymbol interference, nonlinear signal distortion, adjacent channel interference, modem defects, etc. SSM models of the satellite channel with differential and convolutional codes are simulated. The results of simulation agree with the experimental data for the actual channels. Analytical methods for determining the performance characteristics of SSM-modeled systems are developed     

A soft-systems methodology approach for product and processinnovation

Soft-systems methodology (SSM) has been recommended as a tool for scientifically evaluating complex environments. One such environment is the development of organizational processes and products delivered by organizations. This paper proposes a methodology for organizational product and process innovation based on SSM theory incorporating quality function deployment (QFD) and IDEF0 (integrated definition functional modelling tool) techniques. The methodology relies on a series of phases that elicit information from complex and amorphous real-world practices, processes, and information, and develops models of these systems. The paper illustrates the utility of the methodology and tools using a case example: development of an organizational decision support system used to justify strategically important technologies for the enterprise     

Signal extraction with frequency arrangement (SEFA) andsuperimposed subcarrier modulation (SSM) schemes in fiber-orientedwireless access systems

This paper describes technologies for the improvement of the noise and nonlinearity characteristics of wireless access systems based on subcarrier optical transmission. First, the superimposed subcarrier modulation (SSM) technique is proposed to reduce the reflection noise induced by fiber connectors. We experimentally confirmed that the proposed method reduces this noise, and we found the optimum frequency and power of the superimposed subcarrier, SSM design methodology is also presented, taking both noise and nonlinearity into account. Second, signal extraction with frequency arrangement (SEFA) is proposed for increasing the optical modulation index, because the third-order intermodulation distortion (IM3) produced by the optical devices can be ignored. Simulation results show that SEFA greatly improves the transmission performance in terms of both the ratio of desired-to-undesired signal and the receiver sensitivity under Rayleigh fading. Combining the SSM and SEFA is also considered, for use with FP-LDs or DFB-LDs. These technologies will enable cheaper FP-LD implementation in fiber-oriented wireless access systems     

Comparison of Memory Chip Organizations vs Reliability in Virtual Memories

Random access memory organizations typically are chosen for maximum reliability, based on the operation of the memory box itself without concern for the remainder of the computing system. This had led to widespread use of the 1-bit-per-chip, or related organization which uses error correcting codes to minimize the effects of failures occurring in some basic unit such as a word or double word (32 to 64 bits). Such memory boxes are used quite commonly in paged virtual memory systems where the unit for protection is really a page (4K bytes), or in a cache where the unit for protection is a block (32 to 128 bytes), not a double word. With typical high density memory chips and typical ranges of failure rates, the 1-bit-per-chip organization can often maximize page failures in a virtual memory system. For typical cases, a paged virtual memory using a page-per-chip organization can substantially improve reliability, and is potentially far superior to other organizations. This paper first describes the fundamental considerations of organization for memory systems and demonstrates the underlying problems with a simplified case. Then the reliability in terms of lost pages per megabyte due to hard failures over any time period is analyzed for a paged virtual memory organized in both ways. Normalized curves give the lost pages per Mbyte as a function of failure rate and accumulated time. Assuming reasonable failure rates can be achieved, the page-per-chip organization can be 10 to 20 times more reliable than a 1-bit-per-chip scheme.     

PIn. II. Comparative evaluation of SSM/I and TMI precipitable waterestimate for the Mediterranean Sea

For pt.I see ibid., vol.39, no.12, p.2566-74 (2001). To estimate integrated precipitable water vapor along with liquid water path and water vapor effective profile (i.e. standard atmospheric profile approximation), utilizing the Special Sensor Microwave/Imager (SSM/I) and Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) radiometers, an operative procedure was developed and assessed. This procedure is based on a fast nonlinear physical inversion algorithm (PIn) developed by the authors. A large data set of near-coincident TMI and SSM/I data acquisitions were collected and used to supply the procedure. Retrieved parameters were compared against retrievals achieved with well-accepted statistical algorithms, and consistency between TMI and SSM/I retrievals was confirmed. As far as TMI and SSM/I precipitable water retrieving consistency is concerned, this research revealed a linear relationship up to 20 kg/m² and a general overestimate of TMI retrieving, for higher values. A new algorithm for obtaining integrated precipitable water from TMI brightness temperatures was introduced and the goodness of its accuracy was reported. The procedure proved to be reliable and portable and its integrated precipitable water vapor retrieving was assessed to be as accurate as the best radiometric retrieving algorithms, reported in literature. For SSM/I data, developed-procedure liquid water path estimates seemed to be in good agreement with statistical retrievals. Eventually the procedure provided effective water vapor vertical profiles which belong to a deterministic distribution area characterized by an upper and lower limit; it was confirmed that SSM/I and TMI vertical profile distribution areas mainly overlap even if they are characterized by different sensitivities to profile parameters     

Measurement of oceanic wind vector using satellite microwaveradiometers

The possibility of retrieving both wind speed and direction from microwave radiometer measurements of the ocean is studied using Special Sensor Microwave/Imager (SSM/I) measurements collocated with buoy reports from the National Data Buoy Center (NDBC). A physically based algorithm is used to retrieve the wind speed. The RMS difference between the SSM/I and buoy wind speed is 1.6 m/s for 3321 comparisons. It is found that the SSM/I minus buoy wind speed difference is correlated with wind direction. When this wind direction signal is removed, the RMS difference between the SSM/I and buoy winds reduces to 1.3 m/s. The wind direction signal is used to make global, low-resolution maps of the monthly mean oceanic vector. The wind direction sensing capability of a prospective two-look satellite radiometer is also processed     

Lightweight PIM—a new multicast routing protocol

Multicast holds out great promise to significantly reduce network load between the sender and a group of receivers by delivering a single data stream to many recipients. Protocol‐Independent Multicast (PIM) Sparse Mode and its variant PIM Source‐Specific Multicast (PIM SSM) are the most widely used IP multicast protocols. Protocols offer low latency data delivery capabilities with relatively high reliability and are characterized by large signaling overhead. This article presents a new multicast protocol, called Lightweight PIM (LPIM), tackling the main problem hindering widespread usage of IP multicast protocols, which is a large volume of signaling traffic. Given the existing usability of the PIM family protocols, a decision was made to base the new LPIM protocol on PIM SSM. The improvements used in LPIM lead to a considerable reduction of state data and signaling messages required by LPIM as compared with PIM SSM. Similar to its predecessor, LPIM is capable of handling large groups, although with significantly lower overhead. Copyright © 2012 John Wiley & Sons, Ltd.