Performance Evaluation of Range Queries in Key Value Stores

Recently there has been a considerable increase in the number of different Key-Value stores, for supporting data storage and applications on the cloud environment. While all these solutions try to offer highly available and scalable services on the cloud, they are significantly different with each other in terms of the architecture and types of the applications, they try to support. Considering three widely-used such systems: Cassandra, HBase and Voldemort; in this paper we compare them in terms of their support for different types of query workloads. We are mainly focused on the range queries. Unlike HBase and Cassandra that have built-in support for range queries, Voldemort does not support this type of queries via its available API. For this matter, practical techniques are presented on top of Voldemort to support range queries. Our performance evaluation is based on mixed query workloads, in the sense that they contain a combination of short and long range queries, beside other types of typical queries on key-value stores such as lookup and update. We show that there are trade-offs in the performance of the selected system and scheme, and the types of the query workloads that can be processed efficiently.     

硬实时系统在强分区约束下的双层分区调度

文中研究了硬实时系统在强分区约束下的双层分区的调度问题,合理建立了强分区约束下的双层分区调度模型,给出了最坏情况下的分区任务集可调度的判定条件.同时,在此基础上,提出了与分区利用率匹配的分区设计方法,导出了该方法下的系统可调度利用率的最小上限.仿真实验表明,在严格实时的条件下,文中提出的方法相对于现有方法更具优越性,并提高了分区可调度利用率的最小上限.     

Toward cost-effective storage provisioning for DBMSs

Data center operators face a bewildering set of choices when considering how to provision resources on machines with complex I/O subsystems. Modern I/O subsystems often have a rich mix of fast, high performing, but expensive SSDs sitting alongside with cheaper but relatively slower (for random accesses) traditional hard disk drives. The data center operators need to determine how to provision the I/O resources for specific workloads so as to abide by existing service level agreements, while minimizing the total operating cost (TOC) of running the workload, where the TOC includes the amortized hardware costs and the run-time energy costs. The focus of this paper is on introducing this new problem of TOC-based storage allocation, cast in a framework that is compatible with traditional DBMS query optimization and query processing architecture. We also present a heuristic-based solution to this problem, called DOT. We have implemented DOT in PostgreSQL, and experiments using TPC-H and TPC-C demonstrate significant TOC reduction by DOT in various settings.     

Organization and Tagging of Blog and News Entries Based on Content Reuse

As their popularity as dynamic platforms for information dissemination and sharing increases, the use of Weblogs (blogs) which track and comment on real world (political, news, entertainment) events is also growing. The success of the blog as a popular medium for information sharing, on the other hand, is also its weakest spot in that there is little support beyond keyword based searches for blog entries. Consequently, there is impending need for navigational support, which can help users relate a large, diverse, and inherently distributed collection of blogosphere. In this paper, we first note that the existence of large degrees of content overlaps in the form of quotation/commentary pairs (as well as content borrowings across media outlets) can be leveraged for tracking the topic development patterns within the blogosphere. Relying on this observation, we first propose focus and flow analysis techniques that rely on reuse detection and focus and flow to help place blog entries into logical organizations. We then show that these implicit or explicit quotations as well as focus analysis could be leveraged to identify the contexts in which entries occur; thus, resulting in more effective tagging. Thus, we propose CDIP (a collection-driven, yet individuality-preserving tagging system) which relies on relationships provided by quotation/reuse detection and semantic-focus analysis to automatically tag the blogs in such a way that, not-only the related blogs share tags, but also individuality of the entries is preserved for discriminating tag-based accesses.     

强实时约束下光纤通道交换网络消息集优化

实时系统应当具有可预期的定时行为,强实时约束要求系统必须满足定时约束和时限要求.以周期性任务模型作为网络负载可以准确地描述强实时约束下的系统性能.本文以光纤通道交换网络强实时约束下的性能研究为背景,根据调度实时消息传输的加权轮转服务规程,提出了特定消息集的优化设计算法,推导并证明了采用该优化算法后系统最坏情况下的可达负载率比不采用该算法时可提高至少一倍.     

Scalable Filtering of Multiple Generalized-Tree-Pattern Queries over XML Streams

An XML publish/subscribe system needs to filter a large number of queries over XML streams. Most existing systems only consider filtering the simple XPath statements. In this paper, we focus on filtering of the more complex Generalized-Tree-Pattern (GTP) queries. Our filtering mechanism is based on a novel Tree-of-Path (TOP) encoding scheme, which compactly represents the path matches for the entire document. First, we show that the TOP encodings can be efficiently produced via a shared bottom-up path matching. Second, with the aid of this TOP encoding, we can 1) achieve polynomial time and space complexity for post processing, 2) avoid redundant predicate evaluations, 3) allow an efficient duplicate-free and merge join-based algorithm for merging multiple encoded path matches and 4) simplify the processing of GTP queries. Overall our approach maximizes the sharing opportunity across queries by exploiting the suffix as well as prefix sharing. At the same time, our TOP encodings allow efficient post processing for GTP queries. Extensive performance studies show that our GFilter solution not only achieves significantly better filtering performance than state-of-the-art algorithms, but also is capable of efficiently filtering the more complex GTP queries.     

两面针红外指纹图谱与抗癌活性的谱效研究

研究两面针的红外指纹图谱与抗肿瘤活性之间的相互关系.基于不同产地的两面针氯仿提取物的红外指纹图谱特征峰强度及其抗肿瘤活性的效果,采用后退法构建以两面针抑制人胃腺癌7901和人宫颈癌Hela两种肿瘤细胞株的谱效模型.所建立数学模型的预测值与实际测量值的偏差率全部在10%以内,说明两面针红外指纹图谱和两面针抗肿瘤活性之间具有相关性.结果表明,生物碱类成分在两面针抗肿瘤药效活性中发挥了重要的作用.     

使用不同置信级训练样本的神经网络学习方法

针对含不同置信级样本的模型拟合问题,该文提出了一种基于神经网络的二次学习方法。文中指出真实模型是实验模型的一种变异,提出逼近真实模型期望值的神经网络,是融合先验样本和真实样本的最佳网络。首先,以先验样本为训练样本进行第1次神经网络学习,并计算取决于硬点信息的软点误差容量区间;然后,同时将先验样本和真实样本作为训练样本,利用软点误差容量区间和硬点误差敏感系数,对神经网络训练过程中输入/目标对的误差进行修改,通过第2次学习得到既能精确拟合真实样本,又能最大化利用先验样本信息的综合网络。与基于知识的神经网络(KBNN)相比,该方法更加简单,可操控性更强并具有更加明确的逻辑意义。     

Four Promising Multimedia Databases and Their Embodiments

Multimedia database systems have become more and more important as the tool to extract and generate additional values from multimedia Contents. In this paper, four multimedia database systems are proposed from the view point of promising contents sources; the Network multimedia databases, the Stream MM database systems, the Library MM database systems, and the Real world MM database systems. Important problems to be solved, i.e., what to do, are also discussed for each databases. Three concrete multimedia systems by authors' research group, are then introduced and discussed as the embodiments of these multimedia systems; (1) the open Global Image Retrieval and Linking System, GIRLS, for mediation WWW data pace as the network MM database systems, (2) the flexible multimedia database platform GOLS, and (3) the higher level authoring system for the Stream MM environments.     

光纤通道交换机在强实时约束下的分组调度

以光纤通道交换网络强实时约束下的性能研究为背景,采用实时通信中的周期性任务模型,提出了负载匹配的加权轮循分组调度,导出了在该方法下网络消息集严格实时的充要条件,以最差情形下强实时的网络可达负载率为性能衡量指标推证了采用该算法的优越性并通过仿真进行了验证.