SYSTEMS ISSUES IN DATA MANAGEMENT AND INFORMATION RETRIEVAL
 
The goal of systems related research under the auspices of IDM is to
develop a set of software and technologies that is driven by the diverse
needs of database users derived from the two ends of a spectrum.
At one end, fueling this need, are large enterprises who provide
Internet and web-related services, E-commerce -- content providers whose
primary need is the scalability, performance, and reliability of
database systems. At the other end we encounter the critical data
management needs  of scientists involved in the process of scientific discovery
whose primary need is functionality and extensibility of systems.
The group recommends that IDM be cognizant that to develop large-scale,
comprehensive systems requires funding significantly above the current average
IDM individual grant level.  
The working group recommends the following funding themes:

THEME 1: Support for Active Digital Libraries (long term IDM research) 
The typical digital library paradigm tends to describe static and passive
repositories of data. We need to extend this paradigm to enable ``active"
libraries where multiple amounts of diverse types of data
(often from non-conventional ``value-added" types of databases) are added
frequently. We envision these systems to be very dynamic in nature and
self-organizing:
	* reconfigurable and extensible meta-data management: new meta-data 
	  structures that can efficiently describe large volumes of 
          multimedia and GIS data.
	* on-the-fly categorization and  summarization of data: 
	  synopsis/abstraction/summarization 
	  of heterogeneous overlapping data sets including estimating , 
          propagating,  and communicating accuracy and uncertainty measures 
          within those datasets.
	* data validation, correction  and update in VLDBs, and  improved 
	  communication of the content of ``active" libraries:
          A significant part of data loading from real-world data 
          (e.g. satellite imagery) consists of data validation and correction 
          for ``dirty'' data.  Systems should improve the automation 
          of these time-consuming tasks.
	* uncertainty management in data and its propagation: For data which 
          have statistical quality measures (e.g. standard error indicators), 
          these measures should be attached to data and propagated by 
          the system as manipulations of the data occur.
	* integration of framework/models and system: research on data models
          and frameworks for system implementation and evaluation. The
          realization of a system can only be successful if it is
          based on a solid framework derived from a careful requirement 
          analysis and performance considerations.

THEME 2: System Support for Multiple-Paradigms (short and long-term IDM 
research, in collaboration with HCI and Visualization Programs):
New applications require combining multiple query and retrieval paradigms.  
Current   query planning and optimization technology typically focuses on
a single paradigm. The challenge is to develop extensible planners,
optimizers and cost models that can handle multiple paradigms.
	* multiple query and acess paradigms: querying can be over text, video, 
          speech, temporal, geo-spatial data or other, to be defined 
	  data types. Accessing and retrieval can be content based, keyword 
          based, category based or others.
	* multiple representations of data: support can be not only for 
          multiple data types and structures, but can also be dynamically 
          reconfigured for efficiency purposes.
	* constraint databases and interfaces for multiple paradigms:
	  Constraint databases provide a  framework in which multiple 
          types of temporal, geographic, CAD or other spatio-temporal 
          and scientific data can be efficiently represented.  
          In this regard, we need to define expressive and efficient
	  query and update languages as well as fast data access techniques, 
          and build user interfaces with data visualization which 
          can be used by a wide class of people, rather than domain 
          specialists only.     
	* multi-lingual IR (e.g. fonts): At least 100 languages are 
          represented on the WWW.  Standard representations for fonts 
          and character sets only exist for a small number of languages, 
          and even some major languages have several representations 
          (e.g. Japanese and Chinese).  A major support issue is whether 
          to support at the client or at the text server end. 

THEME 3: Performance Conscious Systems Building (short and long-term IDM 
research, in collaboration with Networking Division and industry)
	* integration of Web and database services: 
	  Fast routing of user requests/transactions to the the appropriate 
	  data server in a large IP-based network under 
	  heavy traffic conditions is a core issue in developing effective 
	  e-business systems.  Long network latencies will have to addressed 
	  in the presence of a multi-server operations that are globally 
	  dispersed.
	* integration of process and management in databases: Closer 
	  integration of the supported application processes with the 
	  underlying data management capabilities should lead to more 
	  effective information systems.
	* scalability, performance, reliability, and benchmarking of 
	  information infrastructure/systems development should have 
       	  conscious awareness of the performance limitations imposed by 
	  alternative designs. IDM should support benchmarking for performance 
	  measurement of multi-faceted data paradigms.
	* data management on new networking options:
          The emergence of high-speed network (Gigabits/sec) that
          demonstrate Quality-of-Service guarantees coupled with
          ``infinitely'' fast processors and inexpensive active disk
          are changing the way networked databases will be constructed.
          As dynamic fragmentation and automatic migration of data becomes
          a reality, database will become failure-free even when multiple
          sites may not be available.
          Handling of massive data updates and management of data consistency
          issues will have to be re-examined.
	* permeation of database technology in network management:
	  as Network Routers (both at the edge and at the core) start 
	  employing sophisticated techniques for flow and network management
	  they will increasingly rely on DBMS to store historical and real-time
	  information. Similarly, next-generation DNS and LDAP/LDUP will 
	  have a require close interactions between database and networking 
	  researchers.

THEME 4: Underpinning Software Technologies for Data Intensive Systems
(short and long term IDM research activity in collaboration with Industry).
Object-oriented technology has shown great promise and has helped the 
building of robust data services. In light of recent changes, there might be 
research opportunities in building and validating database services using 
Java and XML. Other enabling technologies include efficient database support 
for networking and software techniques for data intensive systems
	* market competitive object oriented persistent object systems:
	  Providing persistence for objects of arbitrary user defined types
	  makes the technology applicable to non-standard applications.
	  Research efforts on more general models and architecture for
	  persistence are needed.
	* full-fledged XML and Java database technology:
	  Java and XML are elements of a major emerging software technology
          and already have an impact in the field.  These two technologies 
	  could be further enhanced by incorporating multiple model and 
	  architecture for persistence as well as transaction technology.