NSF/CAREER: Robust Data-intensive Cluster Programming

设备复杂性和设计多样性的结合指向系统，这些系统将由固有的和可观察到的异构组件构成。我们需要的是一种新的设计原则，它能解释组件固有的不稳定行为，并促进设计和实现健壮和可预测的系统。 强大的应用程序开发所需的基础设施正在一个新的集群编程系统的背景下被称为冰开发。研究沿着沿着四个轴进行：创建完全鲁棒的数据传输机制，探索先进的适应技术，自动化和易用性的新方法，以及开发鲁棒系统的理论基础。这项工作的总体目标是对如何构建“工作良好”的可伸缩集群系统有一个基本的理解，即使一些底层组件不工作。这种真正强大的系统是可预测的，可靠的，可用的，尽管组成部件的波动和故障，也能表现良好，并且不需要人为干预。这种系统的主要优点之一是减轻了行政负担：只需添加新的组件，并充分利用其能力。

The combination of device complexity and design diversity point towardssystems that will be constructed from components that are both inherently andobservably heterogeneous. What is needed is a new design discipline thataccounts for inherently erratic component behavior, and facilitates thedesign and implementation of robust and predictable systems. The necessary infrastructure for robust application development is beingdeveloped within the context of a new system for cluster programming known asIce. Research proceeds along four axes: the creation of fully robustdata-transfer mechanisms, the exploration of advanced techniques foradaptation, new approaches to automation and ease-of-use, and development ofthe theoretical underpinnings of robust systems. The overall goal of this work is to develop a fundamental understanding of howto build scalable clustered systems that ``work well'', even when some of theunderlying components do not. Such truly robust systems are predictable,reliable, and available, perform well in spite of fluctuations and failures ofconstituent components, and require no human intervention. One of the mainadvantages of such systems is the reduction of the burden of administration:new components are simply added and utilized to their full capacity.