Deployment and Future Directions

1
Deployment and Future Directions

• Stanford Digital Library Project
• Presentation byAndreas Paepcke

2
Deployment Modes
Base technology
Services
Knowledge
Dissemination
3
Deployment Examples

• Base technology transfer
• Jylu Earliest (if not first) mobile CORBA ORB
• ILU improvements
• Object exchange model (OEM)
• Services
• Fab (100 users) Feedback-sensitive Web retrieval
  agent
• InterBib Highly interoperable bibliography
  support

4
Deployment Examples (cont.)

• Interoperability experiments
• CMU skims in SenseMaker. CMU accessing InfoBus
• U. of Illinois/U. of Arizona CS Quest in
  SenseMaker
• Berkeley accessing the InfoBus
• Knowledge dissemination
• Three dozen publications over three years
• Weekly digital library seminar
• Database workshop for Bay Area companies
• Presentation to Stanford Forum industrial
  affiliates
• STARTS
• Graduating students
• Many visitors

5
System Evaluation/Deployment Strategy
User Evaluation Phase I
Technical/UI Improvements

• For User Evaluation Phase II

• DLITE/Sensemaker deployment at
• Xerox PARC
• Stanford University Computer Science library
• NASA library

6
User Evaluation Phase I

• Questions being answered
• Usage models understandable?
• Affordances natural?
• Performance sufficient?
• What do users do with the system?
• Evaluation techniques
• Post-usage questionnaires
• Log analysis
• Observation
• Heuristic evaluation
• Interviews

7
Phase I Evaluation Result Sampler

• DLITE interface
• ? User model exciting and understandable
• ? Front/backend architecture helps deployment
• ? Need multi-threading for
• - Perceived performance
• - Resilience to service failure/overload
• SenseMaker
• ? Users understood and used concepts
• ? Need better UI technologies than HTML/JavaScrip

8
Phase I Evaluation Result Sampler (cont.)

• Fab
• ? System performance improves over time as
  hypothesized
• ? Systems profiles accurately predict user
  preferences
• ? Initial from-zero learning phase not tolerable
• Sound-based interfaces
• ? Blind users reactions very encouraging
• ? Multiple speakers in a single interface is
  useful
• ? Relative differences are difficult to use
• ? Attitude/aptitude around sound in interfaces
  very different for blind vs. sighted users

9
Deployment Challenges

• Use of advanced technology (CORBA, Java, )
• Contractual limitations
• Maintenance expectations

10
Deployment Challenges (cont.)
Stanford Student

ilicon Valley
11
Deployment Challenges (cont.)
System features
Language independence
Threads
PassingDictionaries
Pass ByValue
Platform independence
12
Medium Term Testbed Plans

• Infrastructure
• Parallelism
• Increased code mobility
• Track evolving CORBA facilities(e.g.
  Netscapes/Oracles/Sybases native ORBs)
• Metadata
• Value translation
• Proxy status protocol
• STARTS integration

13
Medium Term Testbed Plans (cont.)

• User interface
• Metadata driven query construction
• Event notification
• Additional services, multi-input service
• Query translation
• Approximate translation
• Additional vector space query facilities
• Continue and grow new research efforts

14
Long Term Plans
Transdu c e r s
Value Filtering
Information-BasedCollaboration
Cellular Repository
Perpetual Activity Service
15
Problem Service Robustness
Corruption
Overload
Network partitioning
Death/Malfunction
Perpetual Activity Service
16
Perpetual Activity Service

• Monitor liveness
• Check correctness
• Restart
• State replication
• Request execution guarantees
• Periodic actions
• Query execution
• ...

Service
17
Problem Federating Digital Libraries
Very Large Scale Storage/Retrieval
Distributed expertise, authority, maintenance
(e.g. WWW)
(e.g. NCSTRL)
Broad data mining interests
(e.g. MVD, Backrub)
(e.g. advertising, intelligence, public health,
dating)
Cellular Repositories
18
Cellular Repositories
CORBA Names
CNRI Handles
...
Music
Film
Virtual Reality
...
Perpetual AS
(E.g. LC)
E.g. Weather on maps
...
WWW
Video
Maps
19
Problem Collaboration in Information-Intensive
Environments
(e.g. workflow support, process-based search)
(e.g. who/when/from-where/what,action control in
info webs)
(e.g. link integrity,view maintenance)
Information-Based Collaboration
20
Information-Based View Email
Initiator
What
Participants
When
Origin
Date Wed, 16 Apr 1997 151629 -0700 (PDT) From
David Maluf ltmaluf_at_DB.Stanford.EDUgt Message-Id
lt199704162216.PAA02829_at_Bigeye.Stanford.EDUgt To
dbseminar_at_CS.Stanford.EDU Subject CS545 DB
Seminar
The database group at Stanford invites you to
attend the second Talk of the 1997 Spring Seminar
Series this Friday (April 18) at 315PM.
21
Information-Based View Source Control
Participants
Initiator
What
When
Origin
RCS file /u/testbed/CVSROOT/dldev/src/interbib/bi
bconvert.py,v Working file bibconvert.py total
revisions 30 selected revisions
30 description ---------------------------- revis
ion 1.30 date 1997/04/01 023658 author
paepcke state Exp lines 5 -3 Fixed xlation
of BibTeX \ idiom which inserts a
char. ----------------------------
22
Information Streams
Crawlers
Newsfeeds
Sensors
Email
When Origin What
Action...
Schema Caster

• Workflow
• Action-aware queries
• ...

23
Longer Term Digital ?Physical World
Physical World
OnlineAuctions
InfoExpress
Digital World
OCR
Print OnDemand
FederalExpress
24
Problem Information Efforts Wasted
Access dynamics
Information
linkage
Value Filtering
25
Value Filtering
EnterpriseDatabases
Value filters
Collaborative recommendation, bookmarks, reading
lists, ...Structural information co-location,
link incidence, phrase length, ... Access
statistics access frequency, re-visitation rate,
access origin
Examples Backrub, Sonia
26
Summary
InfoBusServicesImplemented
Improvements
Evaluation Phase I
You R Here
Improvements/deepening
More publicdeployment
Evaluation Phase II
New projects