Click on a theme or a project in the table below for more information.
Project leader:
Prof.dr. Martin Kersten (CWI)
Consortium:
CWI, UU, UT
Total FTE: 2.90 (heads: faculty: 3, PhD: 2)
Key BRICKS publications:
| • |
M. L. Kersten, S. Manegold: "Cracking the Database Store" In: Proceedings of the Biennial Conference on Innovative Data Systems Research (CIDR), pp 213-224, Asilomar, CA, USA, January 2005
|
| • |
Van Heerde, H.J.W. and Anciaux, N.L.G. and Feng, L. and Apers, P.M.G.: "Balancing smartness and privacy for the Ambient Intelligence" In: Proc. of the 1st European Conference on Smart Sensing and Context (EuroSCC 2006)
|
| • |
S. Idreos, M. L. Kersten, S. Manegold: "Database Cracking" In: Proceedings of the Biennial Conference on Innovative Data Systems Research (CIDR), Asilomar, CA, USA, January 2007
|
|
Project IS1: Organic Databases
The current database deployment palette ranges from networked
sensor-based devices to large data/compute Grids. Both extremes
present common challenges for distributed DBMS technology. The
local storage per device/node/site is severely limited compared to
the total data volume being managed and the local processing power
is too limited to handle a high query load. This project
investigates Armada: a novel reference model for a distributed
database architecture to facilitate evolutionary growth.
Participating systems can autonomously decide to take
responsibility in the distributed data management task. The system
adapts to varying workloads and supports dynamic system re-sizing,
e.g. growing and shrinking of the system at large. Armada uses
lineage trails to capture the meta-data and history. Lineage trails
form the basis to direct updates to the proper sites, break queries
into multi-stage plans, and provides a reference point for site
consistency. The lineage trails are managed in a purely distributed
way, each Armada site is responsible for their persistency and
long-term availability. They provide a minimal, but sufficient
basis to handle all distributed query processing tasks.
Smartness and privacy
A special case of organic databases is found in Ambient
Intelligence environments. They offer smart services to users based
on sensors monitoring users' behaviour to fill personal context
histories. Those context histories are stored on
database/information systems, which we consider as honest: they can
be trusted now, but might be subject to attacks in the future.
Making this assumption implies that protecting context histories by
means of access control might be not enough. To reduce the impact
of possible attacks, we propose to use limited retention
techniques. In our approach, we present applications of a degraded
set of data with a retention delay attached to it, which matches
both application requirements and users privacy wishes. Data
degradation can be twofold: the accuracy of context data can be
lowered such that the less privacy sensitive parts are retained,
and context data can be transformed such that only particular
abilities for application remain avail-able. Retention periods can
be specified to trigger irreversible removal of the context data
from the system.
Industrial cooperation
The activities in the project align with ongoing work at Philips
Research lab and Bsik program Smart Surroundings aimed at
development of database infrastructure for ambient home and ambient
care settings.
The MonetDB platform is concurrently
developed in the Bsik program MultimediaN,
where the emphasis is on multimedia search, Regie voor Geo
Informatiesystems, aimed at improved GIS applications,
and with the Dutch Forensics Institute to simplify
digital forensics.
International cooperation
The basic building blocks have been made public in the open-source
community. Thousands of downloads have been reported. Through this
mechanism there is a plethora of short-term interactions to assess
our database technology and feedback on the solutions sought.
Highlights 2004-2006
Research highlights
In the first phase of the project, we developed an initial version
of the reference models for evolving databases and explored the
limits where privacy enforcement can be enacted. The results have
been presented in, e.g. the keynote of EDBT'06.
Economic & societal impact
Any realistic solution for the envisioned Armada and privacy models
requires a mature experimentation platform. For this we extend the
MonetDB/SQL platform with the necessary application interfacing
tools and techniques to deal with extreme large SQL views.
Future work 2007-2009
The experimentation platform nears its completion. This marks a
turning point in the focus of the project from theoretical analysis
to proof and validation.
The challenge ahead calls for novel
techniques to cope with the dynamic nature of an Armada system,
sites may come and go as they please, data may be on the move
continuously, and never become reachable, etc. In all these cases
we need clear convergence criteria and statistical bounds on the
quality of the results.
IS1 Researchers funded by BRICKS
- Prof.dr. M. Kersten (CWI)
- Dr. P.A. Boncz (CWI)
- Dr. S. Manegold (CWI)
- Drs. S. Idreos (CWI)
- Prof.dr. A.P.J.M. Siebes (UU)
- Drs. H. Philippi (UU)
- Prof.dr. P.M.G. Apers (UT)
- Drs. H.J.W van Heerde (UT)
For more information, please refer to the publications and posters of this project.
|
