CSR: Small: Materialized Views over Heterogeneous Structured Data Sources in a Distributed Event Stream Processing Environment

标题：CSR：Small：分布式事件流处理环境中异类结构化数据源上的物化视图调查者：Suzanne W.DietrichInstitution：Arizona State University项目摘要：软件系统正变得越来越复杂，需要在松散耦合的分布式环境中协调异类结构化数据源，并支持处理事件和流数据。一些样本系统包括国土安全、刑事司法、供应链管理、医疗保健和消费者监控。这样的软件系统涉及用于检测事件、监视条件、处理流和查询数据的大量查询表达式。这项研究分析了这些查询表达式之间对不同语言组件定义的结构化数据源之间的依赖关系，以及关系或以数据为中心的XML，以检测作为物化视图候选的公共子表达式。实体化视图时，将存储计算视图的结果，以便后续引用有效地访问实体化视图，从而避免重新计算的成本。这种性能改进在分布式数据源中更为重要。但是，如果实体化视图所依赖的任何数据源已更改，则必须更新该实体化视图。为了避免代价高昂的重新计算，增量视图维护算法使用更改来增量地计算实体化视图的更新。这项研究的一个独特之处在于，在尊重底层松散耦合的异类数据源的原生格式的同时，高效地维护物化视图。使用最先进的商业和开源组件，支持分布式事件流处理框架的原型环境为探索异质、分布式结构化数据上物化视图的识别、规范和增量评估提供了一个研究和评估平台。这种环境还为本科生研究和课程改进提供了共享的基础设施。

Title: CSR: Small: Materialized Views over Heterogeneous Structured Data Sources in a Distributed Event Stream Processing EnvironmentInvestigator: Suzanne W. DietrichInstitution: Arizona State UniversityProposal #: 0915325Project Abstract: Software systems are becoming increasingly complex, requiring the coordination of heterogeneous structured data sources in a loosely-coupled distributed environment with support for handling events and streaming data. Some sample systems include homeland security, criminal justice, supply chain management, health care, and consumer monitoring. Such software systems involve numerous query expressions for detecting events, monitoring conditions, handling streams, and querying data. This research analyzes the dependencies among these query expressions over structured data sources defined in different language components over relational or data-centric XML to detect common subexpressions as candidates for materialized views. When views are materialized, the results of the computed view are stored so that subsequent references efficiently access the materialized view, avoiding the cost of recomputation. This performance improvement is even more critical with distributed data sources. However, the materialized view must be updated if any data source that it depends on has changed. To avoid costly recomputation, an incremental view maintenance algorithm uses the change to incrementally compute updates to the materialized view. A unique aspect of this research is the efficient maintenance of the materialized views while respecting the native format of the underlying loosely-coupled, heterogeneous data sources. Using state-of-the-art commercial and open-source components, a prototype environment that supports a distributed event stream processing framework provides a research and evaluation platform for the exploration of the identification, specification, and incremental evaluation of materialized views over heterogeneous, distributed structured data. This environment also provides a shared infrastructure for undergraduate research and curriculum enhancement.