A novel system architecture for online operational analytics

用于在线运营分析的新颖系统架构

• 批准号: 463355-2014
• 负责人: Moshovos, Andreas
• 金额: $ 12.69万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Strategic Projects - Group
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=660321
• 关键词: novel; system; architecture; online; operational

Database management systems promised not only persistent storage of transactional data but also their fast and timely (ideally real-time) analysis to support business critical applications. However that promise has always been out of reach due to practical constraints in software and hardware. As a result a typical business data infrastructure is divided between a database transactional store, for "live" business data (e.g. sales, leads, and other opportunities) plus an operational data warehouse for analysis and decision-making. The data images in these two datastores are not in sync, typically lagging by weeks or months.** Recent technological advances are finally converging to bring a unified system within reach, including enormous main memories; highly parallel multi-core processors; fast, high-bandwidth storage systems using solid state disks; and widespread availability of field-programmable custom hardware. The proposal explores a new platform---to be designed from the ground up---that will leverage these new technologies in support of sophisticated, real-time analysis of complex and evolving data, generated by mission-critical systems and business processes. The main emphasis will be around software-hardware co-design to reap the benefits of modern hardware technologies and offer increased performance capable of meeting real-time operational analytics goals. The proposed architecture and system will rely on the solid foundations of the relational data model and SQL and offer self-management and power efficiency, while its real-time capabilities bring true convergence of transactional data and operational analytics.** This research will enable businesses to make real-time decisions on strategy and key functions such as marketing, reacting on business KPIs such as sales in a timely fashion. It will enable delivery of new applications in a world that is struggling to understand and manage big data, such as novel advanced sensing applications and real time medical applications. Finally the work will produce highly qualified personnel in an area of strategic importance to the country such as the management and analysis of big data.

数据库管理系统不仅承诺持久存储事务数据，而且还承诺快速及时（理想情况下是实时）分析，以支持业务关键型应用程序。然而，由于软件和硬件的实际限制，这一承诺一直无法实现。因此，一个典型的业务数据基础设施被划分为数据库事务存储（用于“实时”业务数据（例如销售、销售线索和其他机会））和操作数据仓库（用于分析和决策）。这两个数据存储中的数据映像不同步，通常滞后数周或数月。** 最新的技术进步最终汇聚在一起，使统一的系统触手可及，包括巨大的主存储器;高度并行的多核处理器;使用固态磁盘的快速、高带宽存储系统;以及现场可编程定制硬件的广泛可用性。该提案探讨了一个新的平台-将从头开始设计-它将利用这些新技术，支持对关键任务系统和业务流程产生的复杂和不断变化的数据进行精密、实时的分析。主要重点将围绕软硬件协同设计，以获得现代硬件技术的优势，并提供能够满足实时运营分析目标的更高性能。拟议的架构和系统将依赖于关系数据模型和SQL的坚实基础，并提供自我管理和电源效率，同时其实时功能实现了交易数据和运营分析的真正融合。 这项研究将使企业能够对战略和关键职能（如营销）做出实时决策，并及时对销售等业务KPI做出反应。它将在一个难以理解和管理大数据的世界中提供新的应用程序，例如新颖的高级传感应用程序和真实的时间医疗应用程序。最后，这项工作将在对国家具有战略重要性的领域培养高素质的人才，如大数据的管理和分析。