Transcript Discussion Session

Panel (UoI)
DBGlobe IST-2001-32645
3rd Meeting
Athens, November 29, 2002
Future and Emerging
Technologies (FET)
Proactive initiative on:
Global Computing (GC)
The roots of innovation
1. How is DBGlobe different from p2p; publish/subscribe;
p2p: no distinction between client/server – no centralized control
DBGlobe emphasis: Mobility or in general dynamic environment
heterogeneity (more semantics than file look-up) – no
p2p database yet
may be a p2p (or have p2p components)
small devices (thin peers?)
Publish/subscribe: there are similarities
pmos publish their data/services
subscribe?
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
2. What is the architecture:
what are the basic modules of the system (e.g., indexes,
query processors, etc)
Basic modules: Semantic Mapping components
(ontologies, metadata descriptions)
Indexing components (where is what)
DBGlobe Data Stores (store data, metadata, etc)
PMOs Data Stores (store data)
Processing components
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
2. What is the architecture:
where are the service descriptions stored?
distribution/replication/caching
Need to study this
Replicate (or cache) service descriptions +
directories
Copies of pmo data at proxies
Cache (?) intermediate query results
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
3. Engineering issues:
(use of SOAP and other protocols,
how are services defined by a user, etc)
Part of the implementation
Need to see whether we are compatible when it
comes to implementation
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
4. Do we have a set of "basic services", e.g., allowing to just
export all data from a device?
 I think we should
 In general, we may have “data wrappers” generic
services that just list data – select as a service?
 Is it feasible/reasonable to define a set of “generic”
services appropriate for global computing?
 For instance, location-dependent services?
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
5. (important) How do we handle data produced by
“unregistered” devices -> a system for pervasive computing
should support this
Raw data:
Self-descriptive (annotate them)
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
1. How do we handle (define)
mobility
All at various levels
Mobility
Location just one attribute part of the context
Directories (maintain location information)
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
1. How do we handle (define)
dynamic nature of the environment
Support for updates
Adaptability
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
1. How do we handle (define)
large scale distribution
Locality
Ad-hoc databases or communities
Answer a query inside a context
Horizons may be seen as an instance of this: get all answers
within d hops
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
1. How do we handle (define)
small devices
Proxies:
Act in lieu of their pmo(s)
Store (cache) pmo data, implement services that access data on
the pmo, etc
What does this affect?
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
2. Context: how do we define "context", what is it?
Context:
(a) infrastructure (Location is very important)
(b) User profiles/prefernces
Dynamic or static
Define context as a view (query) over data/services
(e.g., annotate a query with context
information and materialize it as an index)
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
3. Uncertainty
support for non-exact/impresize/approximate
answers
(due to scale, frequency of updates)
We have an instance of this at the index level:
get first k answers (closest to the client)
Other examples?
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
4. Data Mining/Knowledge Discovery
How do we extract information from/about the
environment
Not discussed yet
Perhaps query the metadata?
DBGlobe 3rd Meeting
Athens, Nov 29, 2002
DBGlobe IST-2001-32645