NeTS-NBD: Expressive Resource Specification and Discovery in Configurable Networks

新兴的可配置网络，如大规模覆盖、分布式测试床和传感器网络，包括复杂的、时变的可用资源和网络条件集。此外，这种网络上的分布式应用程序对它们希望利用的资源的要求也越来越复杂。例如：识别并启动摄像机以跟踪物体在受监控空间中的预期轨迹；计划网格中的覆盖多播文件传输；或在PlanetLab中嵌入具有特定资源限制的网络实验。PI研究了在必须将特定于应用程序的要求实例化到大规模可配置网络上时，这些设置所面临的常见挑战。现有方法通常不能以适当的细节级别来表征网络条件和应用需求，该细节级别平衡了网络测量量和粗粒度表示的不精确度。PI正在通过两个相辅相成的方面的工作来改善这种状况：(1)研究应用需求和网络条件的两种新的、可调的和通用的形式表示，以及(2)开发用于实现特定网络设置的应用需求的高效算法和技术。PI正在解决新的问题，例如如何在满足目标要求的同时将应用程序高效地映射到可配置网络上，以及如何进行一组足以获得应用程序及其要求的给定嵌入的最小网络测量。PIS预计，拟议的工作将提供一套有用的构建块，可以改进联网应用程序的设计和性能，并有助于为未来网络的设计提供信息。他们的目的是促进研究小组之间关于如何最好地构建适用于当前和未来各种可配置网络的工具包组件的富有成效的讨论。

Emerging configurable networks such as large-scale overlays, distributed testbeds, and sensor networks comprise complex, time-varying sets of available resources and network conditions. Also, distributed applications atop such networks have increasingly complex requirements on resources they wish to utilize. Examples include: identifying and actuating cameras to track an object's expected trajectory through a monitored space; scheduling an overlay multicast file transfer across the Grid; or embedding a network experiment with specific resource constraints into PlanetLab. The PIs study challenges common to these settings that arise when application-specific requirements must be instantiated onto a large-scale configurable network. Existing methods are not generally capable of characterizing network conditions and application requirements at an appropriate level of detail that balances the amount of network measurements and the imprecision of a coarse-grained representation. The PIs are improving this state of affairs by pursuing work along two complimentary dimensions: (1) the study of two novel, formal representations of application requirements and network conditions that are approximate, tunable, and general-purpose, and (2) the development of efficient algorithms and techniques for the realization of application requirements given a particular network setting. The PIs are addressing new questions such as how to efficiently map an application onto a configurable network while satisfying the target requirements, and how to conduct a set of minimal network measurements that are sufficient to obtain a given embedding of an application and its requirements. The PIs anticipate that the proposed work will provide a set of useful building blocks that can improve the design and performance of networked applications, and can help inform the design of future networks. Their intent is to foster productive discussion across research groups about how best to build components of toolkits applicable across a diverse set of current and future configurable networks.