Completed Research
Projects
· BestPeer
·
BestPeer
2.0: a peer-to-peer based data management system which supports both
structured (BATON) and
unstructured with supernodes overlays. It is a one-platform multi-structures
p2p network that is able to support expansion-based routing in unstructured
setting, and DHT-based exact and range-based routing in structured setting. It
supports keyword, similarity and attribute based query retrieval. It is
designed to provide enterprise-quality data processing capability in corporate
networks. It uses JIT and approximate indexing to reduce maintenance cost.
The system is built in lego block manner, and each main concept/component is
tested on PlanetLab before it is incorporated into BestPeer. Related projects: S3(BestPeer 2.0), and BestPeer Pte Ltd
· COSMOS
COMOS (COoperative
and Self-tuning Management Of Streaming data)
focuses on developing algorithms for streaming data. In particular, we
emphasize on self-tuning management.
· UNOdb
In UNOdb (UNcheatable Outsourced
databases), we developed
technologies for authenticating query results. We have developed methods for
SPJ queries, aggregate queries, and spatial queries (window, range, knn).
· GENESIS
GENESIS (GENomE Seqence Indexing and Searching) studies
indexing techniques to speed up sequence and structure database searches.
· StegFS
- StegFS (Steganographic File System) is a prototype file
system that can be used to hide private files so that they are essentially
non-existent to users without access.
· REVIEW
- REVIEW (REal-time VIrtual Environment Walkthrough) is a
prototype walkthrough system that has been designed (and currently under
implementation) for gigabyte virtual environment that does not fit into
the main memory. We are studying methods to reduce (1) I/O cost (using
novel index structure that integrates visibility and spatial information);
(2) CPU cost and (3) GPU (Graphics Pipeline Unit) cost.
· VIPER
VIPER
(VIsual Property-based search Engine for image Retrieval) studies mechanisms to
effectively and efficiently retrieve images from a large image database. For
the former, we have and are currently looking at novel image representation
techniques. The latter addresses access methods to facilitate speedy retrieval
of images. Technologies in this work has been both patented and transferred to Geofoto Pte Ltd!
· DIVA
- This project addresses image and video retrieval for
large collection of multimedia data.
· XENA
- In XENA (XML sEcurity eNforcement Architecture), we are
addressing the issue of designing a scalable access control system for XML
data. Our approach is to integrate access control polices designed by the
XML community with XML storage mechanisms designed by the database
community. The system has the following features. First, it can regulate
access control at a fine graularity (e.g., at the tag level). Second, it
stores XML documents as tables in relational databases. Third, it is
efficient compared to existing systems as it only examines the relevant
data. Fourth, it is scalable as it can handle very large XML documents
that may not fit into the main memory. Finally, it provides very fast
initial response time. For a demo, see here. This project is part of
the Strategic Program on Computer Security funded by the National Science
and Technology Board and the Ministry of Education.
· GO
GREEN!
- In this project, we explore techniques that are
"environmentally-friendly" (our environment refers to the system
environment). Essentially, we attempt to do so with the following three R's:
- Reuse.
Techniques that reuses available data and/or structures should be
designed. For example, we can cache intermediate or final results of
queries and reuse them for subsequent queries. This has the advantage of
speeding up the query evaluation of subsequent queries.
- Recyle.
Techniques that recyles existing data or structures are preferred. For
example, reorganization of a dataset may recycle existing indexes, rather
than rebuilding an index from scratch.
- Reduce.
Techniques that minimizes the processing cost, and energy (for portable
equipments) are highly desirable. Designing efficient algorithms has
always been an active area of research. Power saving techniques are
becoming increasingly important for portable devices.
· MERLIN
- In a digital marketplace (DigiMart), only
digital good are traded in the market. The nature of digital goods raises
a bunch of problems which has new implications for research in different
disciplines. In particular, we are interested in the efficient evaluation
of queries that provides fast first answers to the users without
sacrificing on the efficiency of the system. New mechanisms are being
explored to deal with slow and unpredictable delays in the data arrival
rates. These include the idea of information foliaging, new symmetric
join processing, new dynamic query optimization techniques, and
the exploitation of redundant/overlapping information repositories.
· CRAM
- This project explores energy efficient solutions to
information dissemination in wireless environments.
· MOSAIC
- This project studies indexing mechanism to facilitate
speedy retrieval of images from a large image database.
· SWIFT
- This project aims to provide users with fast access to
information. This is achieved by minimizing initial response time (through
the phenomenal of ``getting some answers quickly, and then more later''
and minimizing overall processing time using novel query processing and
optimization techniques. Click here for more information and a demo.
Another approach adopted is the online evaluation of complex queries
involving aggregates allows the tradeoff between fast initial response
time and accuracy of answers.