Claremont Report on Database Research: Research Directions (Le Gruenwald)
3 likes898 views
The document discusses future directions for database research, highlighting the need for formal data semantics, integration of diverse databases, and addressing usability for varied users. It emphasizes the importance of mobile database applications and the necessity for multi-disciplinary collaboration in database research. Additionally, it identifies challenges in database performance evaluation and integration with programming languages, calling for innovative solutions and improved methodologies.
Claremont Report on Database Research: Research Directions (Le Gruenwald)
- 1. What is to be done? The Future of Database Research Le Gruenwald National Science Foundation Presented to 2008 Database Self-Assessment Submit May 29-30, 2008
- 2. Topics to Consider Formal Data Semantics Graph Database Human-Centered Database Computing Multi-disciplinary Database Research Mobile Database Database Performance Evaluation
- 3. Formal Data Semantics For most of the past 40 years DB community has largely ignored most issues concerning data semantics, even such basic matters as measurement units. Nearly all DB systems today lack formal specification of data semantics. This issue is of increasing importance as we attempt to integrate more diverse databases. Need to provide formal data semantics (metadata), e.g., logic – but which logic? DL, FOL, sorted, ... Need ability to integrate and query data semantics. Increasing demands for integrated DB retrieval and inference.
- 4. Graph Database Big demand: transportation, bio, social networks E.g. perform disjunctive queries over different relationship types E.g. find the shortest path from point A to point B Need a flexible data model and query language
- 5. Human-Centered Database Computing Need to accommodate different types of users Usability studies Visualization
- 6. Multi-disciplinary DB Research Need to reach out to other disciplines: What are the innovative uses of existing DB research results that enable transformative research in other disciplines? What transformative DB research would be derived from the needs of other disciplines? Major DB conferences and journals need to embrace multi-disciplinary DB research
- 7. Mobile Database Increasing demand for mobile applications (including mobile sensor applications) Issues: mobility, disconnection, energy limitation, etc. More activities in this area in Europe and Japan than in the U.S. Major DB conferences need to embrace mobile database research Can energy-aware mobile DB research be extended to achieve GREEN DB for static environments?
- 8. Database Performance Evaluation Many of current DB research evaluation plans include: Performing simulation experiments using Synthetic datasets Real-life datasets Benchmark datasets (not always available) Making some generalized conclusions without regards to statistical relevance Too ad-hoc, lack of science -> Need a more credible evaluation approach
- 9. THANK YOU!
- 10. Extra Slides for additional topics
- 11. Data Models for Vector Fields Vector fields occur in many scientific, engineering applications: Computational fluid dynamics: weather, climate, oceanography, airplane design, wind turbine design and placement, finite element modeling, .... Relational model is largely useless Attempts: Fiber Bundle Data Model (lloyd Treinish, ibm walson david butler, limit point), Vector Bundle Data Model (eddie saek, richard Muntz, ucla ...) /*restricive fiber with map from mesh to vector space from one end to another end */ Need data models, query languages, ... Need interpolation
- 12. Shape Based Retrieval Applications: Part retrieval, protein docking, protein-ligand binding, drug design, archeology, airplane crash reconstruction, ... Need invariant shape descriptions w.r.t. translation and rotation Need efficient representations and query processing Need methods for “compliant” shape matching (docking) /* mating */
- 13. Impedance Mismatch Between Programming Languages and DBMSs Longstanding problem of integration of queries into programs Generally poor support by programming languages OODBMSs failed Latest effort: Microsoft Linq Remains, open, difficult problem See related work on XDUCE, CDUCE
- 14. Very Large Data Integration Data Integration / DB Federation over large numbers of DB (100's or 1000's) remains unsolved problem Increasing important for bioinformatics, intelligence, e-commerce, ... Need better metadata, better tools, new approaches ??