If you want to discuss MQP ideas please stop by my office (238 Fuller Labs) or email me at rundenst@cs.wpi.edu for an appointment.
This year I'm particularly interested in projects related to the newly emerging area of "stream and event monitoring systems". Examples of possible projects include :
Advanced Services and Applications for Stream Processing Technology
Project Description: Continuous high-volumne data streams can be found everywhere, including vehicle navigation tracking systems, sensors for health applications, earth observation systems, stock market, and more. The goal of our larger project effort here at WPI involving database and software engineering faculty, called CAPE (Continous Adaptive Processing Engine), is to develop generic technology to support such stream monitoring applications. We have a core engine in place. Several challenging new MQP projects can now be undertaken within that platform.
One example project could be to explore strategies for spilling (and unspilling) data of different query operators temporarily to disk under heavy duress (high input arrival rates of some streams and/or reduction in CPU or memory resources in the system). Techniques need to be proposed, data structures and algorithms implemented, and policies for synchronization between operators need to be realized for improved performance even under spill.
Heterogenous streams of events, such as from RFID readers, may need to be quickly scanned for detecting patterns of interest, such as an item leaving the store without having been paid for, etc. Query operators utilizing automaton/stacks would be developed, and advanced methods to handle such pattern matching services even under out-of-order input streams are targeted.
RFID technology has now become a reality due to its inexpensive costs, and it is been increasingly employed in a large variety of environments, from airports to warehouse supply chains. This project would target to install such technology on campus, and to develop a simple data collection and tracking application to support one function, such as misplacement of books, or counting of number of people using a facility, etc. Accuracy in data processing and reasoning under uncertainty, such as missed readings or duplicate readings, as well as exploitation of domain knowledge, such as sizes of rooms and typical speed in traversing a given space, could be considered to make the application smart.
We have this year acquired a brand new 20-node PC cluster for 100k by NSF for developing stream processing software and applying this technology to a variety of applications. Thus, MQP projects surrounding this technology are highly encouraged. Some sample projects one could conduct include :
Explore, implement and test algorithms for allocating heavy query operators in a query plan across the compute network. That is,to explore strategies for distributing computations across clusters of machines pushing the query filters closer to the remote data stream generators. Further, design different strategies and software protocols for relocating at run-time operators across machines, or possibly even reoptimize a query plan and then migrate to new plan shape at run-time. Java and software engineering skills are desired. DB courses a plus.
DCAPE assumes that all servers are up and the network is always reliable. This project would look at graceful methods for recovering or preparing for failures of the different components of the stream engine, be it a machine, a query engine, a connection, or an end application. Implementation and testing are needed to validate the proposed ideas for achieving recovery under failure.
Right now, the system contains some initial light-weight yet brittle solution for monitoring the performance of each of the query engines distributed across the cluster. The project would aim to develop an infrastructure to simplify installation of the ECAPE system on the cluster, run-time experimentation, smooth shut-down, and collection of results into automatically generated result charts for analysis by the students working with the different enhancements of the stream engine.
Look at on-going larger research projects of the WPI Database Systems Research Lab at the " following location", any of which you can get involved in, if you are interested.
If you have questions and/or are interested in any of these projects, please contact Prof. Rundensteiner at rundenst@cs.wpi.edu to find out more.