DBIS-TOOLKIT

DBIS-TOOLKIT

Adaptable Middleware For Large Scale Data Delivery

System Support for Prototyping and Developing

Dissemination-Based Information Systems

DBIS Applications

Data Dissemination:

Stock and sports tickers

Traffic information systems

Personalized news

Entertainment delivery

Common Properties:

Tremendous scale

Significant overlap in user data needs

Asymmetric data flow from sources to consumers

Problems with Current Architectures:

Huge delays in getting answers

High resource consumption

Scalability limitations

Inability to predict performance

DBIS Framework

Properties:

Multiple Data Delivery Options

Network Transparency

Three node types:

Data Sources

Information Brokers

Clients

Instrumentation/Visualization Support

Network Transparency:

network.gif (2507 bytes)

Example:

Push-based Proxy Server for Pull-based Web clients

proxy.gif (3785 bytes)

Delivery Options

Data Delivery:

Three Orthogonal Dimensions:
Push vs. Pull
Aperiodic vs. Periodic
Unicast vs. Broadcast/Multicast

Main Features

Allows dynamic mixing-and-matching of data delivery options within a single application.

Exploitation of network transparency through the use of multiple levels of Information Brokers.

Flexible, real-time performance monitoring.

Broker-based caching and profile matching.

RxW Algorithm for broadcast scheduling and data staging.

Toolkit in Action -- Weather Map Dissemination

dbis.jpg (92448 bytes)

Project Members

Stan Zdonik

For more information:

http://www.cs.umd.edu/projects/bdisk

Sponsors

darpa.gif (6867 bytes) nsf.gif (5975 bytes)