The distributed computing and communication infrastructure provides the underlying computational architecture that links, operates, and coordinates the whole DCAS system and all its networks. The Distributing Research Thrust is concerned with such key functions as data movement, data storage, archival processing, and optimizing how all the system resources are dynamically employed to meet end user data needs.

The Distributing Thrust will focus on several critical issues:

• To develop the end user and application-driven adaptive resource allocation and optimization algorithms for DCAS systems that include hazardous wind environments and rain.
• To develop an integrative, distributed system sense-and-response software architecture that can meet the adaptivity and timeliness requirements dictated by end user needs, and the intrinsic properties of the environment being sensed. Key components of this architecture include computing, networking, and storage.
• To develop the architecture and protocols for remote-sensing networks in resource-poor environments.

In order to meet these goals, we are undertaking a broad set of research projects, ranging from near-term to long-term and from applied development to theoretical analyses and modeling in the sub-disciplines of computing, networking, and storage. Several of our projects lie at the intersection of two or more of these sub-disciplines, as shown in this diagram. To see the results of our research, please go to the Distributing Thrust publications webpage.

Research Accomplishments

Overall Software System Architecture
The Meteorological Command and Control (MC&C) performs the system's main control loop - ingesting data from remote radars, identifying meteorological features in this data, making features available for presentation to end users, and optimizing the configuration of each radar's future scan strategy based on detected features and end user requirements. We have "closed the loop" with a live MC&C controlling the Massachusetts Test Bed radar, and benchmarked the MC&C component performance. Using the DCAS emulator, we have also benchmarked performance in an emulated 4-node CASA DCAS network, demonstrating the ability of the software system architecture and the provisioned computing infrastructure to meet IP1's computational and timing requirements.

Optimizing the Use of Radar Resource
The goals of this project are to optimize the targeting of a set of overlapping and collaborating radars to direct their scanning when and where end user needs are greatest. CASA researchers from the End User, Sensing, Distributing, and Predicting thrusts have jointly refind the utility function that characterizes the value of a particular sector scan to the end-to-end system. Using the DCAS emulator, we have begun a performance evaluation of sit-and-spin versus targeted-sector scanning, obtaining initial results illustrating the higher utility gained by adaptive sector-scanning over sit-and-spin. We have initiated projects on optimal wind field retrieval and on a comparison of myopic versus multi-time-step optimization, with promising results.

TSAR: A 2-Tier Sensor Storage Architecture Using Interval Skip Graphs
We have designed and implemented a tiered storage architecture for tiered architectures like CASA that comprise both the radars at the edge as well as a network operations center (NOC) that aggregate data from multiple radars. TSAR is a fundamentally different storage architecture that envisions separation of data from meta-data by employing local archiving at the edge radars and distributed indexing at the NOC. At the NOCs, TSAR employs a novel multi-resolution ordered distributed index structure, the Interval Skip Graph, for efficiently supporting spatio-temporal and value queries. At the edge radars, TSAR supports energy-wave adaptive summarization that can tradeoff the cost of transmitting metadata against the overhead of false hits resulting from querying a coarse-grain index.

Data Transport Protocols for Radar Networks
We are working on the design, analysis, and implementation of several new data transport protocols for transmitting radar data between endpoints in different scenarios: the TRABOL protocol provides for real-time, application sensitiv, and congestion-controlled data transfer; many-to-one, and one-many overlay routing approaches to more effectively utilize the aggregate bandwidth from the radars to the SOCC, and among consumers of radar data; and an application-level TCP-nice-like background transfer protocol for low priority radar data.

Networking in Resource-Challenged Environments
We have developed and analyzed a distributed gradient-based algorithm for optimal joint allocation of energy between sensing and communication at each node to maximize the aggregate amount of data received at the sink. This research is targeted toward the Puerto Rico Off the Grid test bed, with wireless communication links and harvested solar power. We have proven that the algorithm converges to the point of maximum system utility, and quantitatively demonstrated the energy balance achieved in a simulated model of the test bed using small boat radars and wireless cards that will be used in the Puerto Rico test bed.

(back to top)