Stratosphere Data and Processing Model, its Optimization and Parallelization

平流层研究单位内的项目A的目标是在大规模并行、容错、自适应计算体系结构上优化和并行化声明性数据流程序。作为整个研究单位的基础，该项目将协调规划模式的设计，其中文本数据和不确定性是一等公民，可以具体说明和以最佳方式执行数据密集型操作，例如信息提取和数据清理操作符。该项目将研究如何通过提供性能度量、优化算法和并行数据流编程的抽象来有效和高效地优化这些数据流程序的健壮性和并行化执行。该项目还将通过分析和实施来自气候研究的真实世界用例以及通过评估和基准系统性能来展示平流层方法的整体有效性。

The goal of Project A within the Stratosphere research unit is to optimize and parallelize the declarative data flow programs on a massively parallel, fault-tolerant, adaptive computing architecture. As foundation for the overall research unit, the project will coordinate the design of a programming model, where text data and uncertainty are first class citizens, and data-intensive operations, e.g., information extraction, and data cleansing operators, can be specified and optimally executed. The project will research how to effectively and efficiently optimize the robustness and parallelize the execution of these data flow programs by providing performance metrics, optimization algorithms, and an abstraction for parallel data flow programming. The project will also demonstrate the overall effectiveness of the Stratosphere approach by analyzing and implementing a real-world use-case from climate research and by evaluating and benchmarking the system performance.