Distributed Real-Time Databases



BeeHive Project

The BeeHive Project is developing technology required for global, object-oriented real-time databases with emphasis on adding value along four dimensions: real-time, fault tolerance, security, and Quality of Service (QoS). Real-time databases exist when transactions have deadlines and where data is valid only for a period of time, e.g., it could be data from sensors or derived from sensors. Applications include Internet services, defense applications, and smart spaces.

Our work includes a single node (2-processor) BeeHive real-time database testbed as well as several simulation systems. The BeeHive testbed is built on the Wisconsin Shore storage manager and includes admission control and various real-time database protocols such as earliest data deadline first and forced delay. Experiments are currently being performed to study the performance of these protocols and the value of admission control in real-time database applications. The results in this area are the first reported real-time database results of this type from an actual testbed implementation. Previous work was all done by simulation.

Many enterprises will require the ability to dynamically create virtual databases with the above properties AND connect to sensors and actuators. BeeHive is investigating many research issues related to supporting this vision. As part of this we are also studying how to support various types of Quality-of-Service (QoS) for both transaction timeliness and data freshness in real-time databases. We are also developing feedback controllers to enforce different guarantees for different classes of service. This is sometimes called a differentiated service model.

The project has also developed real-time logging and recovery protocols and techniques to use time signatures for detecting data intrusions.

Our project is also investigation the development of data services in large scale sensornets. Here many of the same issues investigated on the BeeHive testbed appear, but there are additional severe constraints due to the small size of the devices, the power requirements of the devices, and the high failure rates in communications and the devices themselves. We currently have results on a data caching algorithm for sensornets and are implementing our solutions on a wireless network of motes (developed at Berkeley).

In this project, in the past, we have interacted with Unisys corporation and Kyung Hee University. Currently, we are funded by the National Science Foundation.


Principal Investigators

Jack Stankovic
Sang Son

Current Graduate Students

Binjia Jiao
Yuan Wei
Ying Lin

Recent Long Term Research Visitors

SuHee Kim, Korea
Sooyeon Kim, Korea
Pierre Aussourd, France
LihChyun Shu, Taiwan
Victor Lee, Hong Kong
Jorgen Hansson, Sweden

Prof. Park, Seoul National University

Selected Publications