III: Small: A User-centric Architecture for Ad-hoc Analytical Processing

考虑到使用基于云的基础设施进行交互式数据探索的需求不断增长，PI提出了一种以用户为中心的架构，专门为经济高效的实时分析处理而设计。建议的架构将使用答案近似技术和分布式内存查询评估，以提供一个流体的数据交互体验的最终用户。以用户为中心的交互范式有可能彻底改变未来的数据处理平台，并迅速加速数据驱动的发现和决策，如商业管理，生物信息学，天文学，金融等领域。PI提出了一种新的交互式分面探索范式，用于分析多维数据立方体。通过紧密集成在一个分布式的，主存交互式查询引擎的数据立方体的分面探索，该项目研究了三个互补的途径：面向交互的立方体探索模型，交互式主存在线分析处理（OLAP）的查询评估策略，和分布式架构的并行分析，贸易的准确性，以满足用户定义的响应时间的限制。PI将研究一种独特的技术组合，如推测执行、共享扫描和采样，以服务于以用户为中心的OLAP工作负载。分面探索模型和底层数据库引擎原语的同时设计允许以直观和交互的方式执行数据探索任务，同时利用现代硬件功能来提高性能。通过关注用户，该项目将重新考虑数据库体系结构的各个方面，以便交互式地探索大型数据库。有关详细信息，请参阅项目网站：http://go.osu.edu/userolap

Given the growing demand for interactive data exploration using cloud-based infrastructure, the PIs propose a user-centric architecture designed specifically for cost-effective, ad-hoc, real-time analytical processing. The proposed architecture will use answer approximation techniques and distributed in-memory query evaluation to offer a fluid data interaction experience to the end user. The user-centric interaction paradigm has the potential to revolutionize future data processing platforms, and rapidly accelerate data-driven discovery and decision making in areas such as business management, bioinformatics, astronomy, finance, and more.The PIs propose a novel interactive faceted exploration paradigm for analyzing multi-dimensional data cubes. By tightly integrating faceted exploration of data cubes within a distributed, main-memory interactive query engine, this project investigates three complementary avenues: cube exploration models geared towards interaction, query evaluation strategies for interactive main-memory online analytical processing (OLAP), and distributed architectures for parallel analysis that trade accuracy to meet user-defined response time constraints. The PIs will investigate a unique combination of techniques such as speculative execution, shared scans and sampling to serve user-centric aspects of OLAP workloads. The simultaneous design of faceted exploration models and underlying database engine primitives allows data exploration tasks to be performed in an intuitive and interactive manner, while leveraging modern hardware features for performance. By focusing on the user, this project will rethink aspects of database architectures for interactively exploring large datasets.For further information, see the project web site at: http://go.osu.edu/userolap