"HMI? Case Studies on Scientific Research at MIT" Webinar July 2009

advertisement
DataSpace Overview – NSF Site Visit
8 February 2010
(v8)
1
Vision
• “To bring the dramatic benefits of the Web to scientists –
comparable to the benefits the Web has had to commerce and
other areas”
• . . . Not just in the impact to science, but also a similar
distributed federated ecosystem for:
– Technology Infrastructure
– Organizational Responsibilities
• View: Research data-generating institutions and their libraries
should play an active role in curating their researchers’ data
– Financial and Technical Sustainability
– Openness: 3rd party extension and Open Source development
• Support research across all domains, but initially:
– Neuroscience
– Biological Oceanography
2
Basic Data User Services:
Discovery, Quality, Conversion, Integration
Data Curation Services:
Process, Catalog, Annotate, Preserve
DataSpace
High-Level
Architecture
Additional Data User
Services :
• Data Analytics
• Data Visualization
Distributed Data Management Services:
Security, Replication, Administration
Policy Management, Workflow Services
Global Network (Web)
DataSpace
Services
Local Network
...
...
Interface to
Legacy Scientific
Data Repositories
Metadata
Repository
for Scientific Data
Other USA
Nodes
3rd par
Multiple Scientific Data Repositories
(DataSpace Native Architecture)
Basic Workflow
Scientist
Provides
data,
preliminary
metadata
International
Nodes
3rd Party Specialized
Data Services
MIT Node
Curator
Process and
ingests data,
complete
metadata, and
policies (e.g.
retention)
DataSpace
User
Searches (meta)data,
accesses/integrates
data, analyzes/visualizes
data (via DataSpace data
services or 3rd party 3
data services)
Initial Scientific Domains Chosen
• Neuroscience and Biological Oceanography
– Sciences with complex interdisciplinary sub-domains
– Different and diverse types of scientific data
• Though some aspects of overlap (genetic data)
– Faced with challenges related to
• Data expression, encoding, sharing, integration, visualizing,
and preserving
• Difficult to perform research that crosses sub-domains or
requires multi-source data
– Can build on existing collections and collaborations
• But must also address technical, social, and legal issues
• Will bring in additional domains over time
4
DataSpace Organizational Structure (preliminary)
DataSpace Advisory Board
(external, international,
10-15 members)
Outside Parners
DataNet Partners
Other
Cyberinfrastructure
Partners
Management Board
(PIs & Senior
Personnel)
Other Sectors
(e.g. finance,
pharma, health
care, insurance)
Research &
Prototyping
• Data Protection, Security
• Distributed Policy Management
• Data Analysis, Visualization
• Data Analytics
• Data Sharing Policy and Legal Advise
• Data Quality
• Data Semantics, Discovery
• Data Interoperability, Integration
• Data Storage Architecture
• Data Curation Workflows
DataSpace PI
DataSpace Project
Director
Sloan School of
Management Finance
& Administration
DataSpace Business
Development Team
(DBDMT)
Development &
Operations
Marketing &
Outreach
• Cyberinfrastructure Architecture
• Software Design & Development
• Infrastructure Planning
• Data Curation Operations
• Technology Operations
• Service Modeling
• Business Modeling
• Communication
• Coordination
• User Needs Assessment
• Usability and Feedback
• Public outreach
(‘citizen science’)
• Educational outreach
• Scholarly Publishing outreach
Some Key Goals for First Year
• Complete hiring and staffing
• Design and development of DataSpace v1 (Interim architecture)
– Build on existing software base (DSpace, Fedora)
– Addition of initial DataSpace middleware
• Ingest of initial Neuroscience and Biological Oceanography data
– Selection/development of ontologies
– Recording of metadata (including preservation policies, etc.)
• Establish operational DataSpace v1
– Service models defined with partner nodes
• Design of DataSpace long-term architecture
– Initial results from research groups for v2
• Initial results of Business Development Management Team
• Educational and Outreach efforts (i-schools, OCW)
6
Sustainability Approach
• Core to Financial Sustainability
– Provide maximum value to science
– Minimize cost to any one organization by broad distribution
• Can actually reduce costs by eliminating duplication and inefficiencies
– Build on the long-standing role and sustainability of libraries
– Follows Web/Internet value (to both large and small orgs)
• Worldwide infrastructure, costs widely shared
• Technological Sustainability
– Open Source software, multiple implementations possible, and
encourage 3rd party augmentation
– Participation of commercial technology company partners
• Some Resources: DataSpace Federation, Partner experiences,
Business Development Management Team (working with MIT
Entrepreneurship Center, E&I students, etc.)
7
Some Key Features of the DataSpace Proposal
• Distributed federated infrastructure for accessibility & long-term preservation
– Address privacy, property and data rights, etc. with legal and policy framework
•
•
•
•
•
•
•
•
•
Builds on successful Dspace/Fedora platform
Proposes new top-level internet domain (".arc")
Addresses need for “temporal semantics” and other advanced metadata
Risk mitigation: Research risk: Personnel with extensive experience.
Operational risk and sustainability: Distributed design and federated approach.
Public/Private Partnership: Corporate partners help build more sustainable
ecosystem and ensure sustainability, MIT Entrepreneurship Center, etc.
Expert Advisory Board: Diverse fields (i.e. science, law, business, technology,
libraries, and digital preservation) advise and promote the project
Advances scholarly communications through data/publication integration
Advances educational technology through data/courseware integration
Outreach to minority and pre-college student, underserved small and medium
research groups.
DataSpace will be a truly transformational project
8
Multi-disciplinary team of Principal Investigators
• Hal Abelson, MIT Computer Science & Artificial
Intelligence Laboratory (CSAIL)
• Ed DeLong, MIT Departments of Civil and
Environment Engineering and Biological Engineering
• John Gabrieli, MIT Department of Brain and
Cognitive Sciences
• Stuart Madnick, MIT Sloan School of Management &
School of Engineering
• MacKenzie Smith, MIT Libraries
• Marilyn T. Smith, Director, MIT Information Systems
& Technology (IS&T) [replaces Jerry Grochow]
9
Diverse and Experienced Senior Personnel
• Timothy Berners-Lee
(W3C, WSRI)
• Alon Halevy (Google)
• Geneva Henry (Rice
University)
• Mei Hsu (HP)
• David Karger (MIT)
• Michele Kimpton (DSpace
Foundation)
• Thomas Malone (MIT)
• Dejan Milojicic (HP)
[replaces John Erickson]
• Joe Pato (HP)
• Terry Reese (Oregon State
University)
• Michael Siegel (MIT)
• Stephen Todd (EMC)
• Tyler Walters (Georgia
Tech)
• Danny Weitzner (W3C,
WSRI)
• Steve White (Microsoft)
[addition to team]
• John Wilbanks (Science
Commons)
• Wei Lee Woon (MIST, Abu
Dhabi)
10
Advisory Board
• Christine L. Borgman, Department of Information Studies, Graduate
School of Education and Information Science, UCLA
• Randy Buckner, Psychology, Harvard University
• Scott Doney, Marine Chemistry & Geochemistry, Woods Hole
Oceanographic Institution
• Keith Jeffery, European Research Consortium of Informatics and
Mathematics (ERCIM) and UK Rutherford Appleton Laboratory
• Liz Lyon, UKOLN and UK Digital Curation Centre
• Ed Roberts, Management of Technology, MIT Sloan School of
Management and MIT Entrepreneurship Center
• Pam Samuelson, School of Information and School of Law , UC Berkeley
• Dan Schutzer, Financial Services Technical Consortium (FSTC)
• Andrew Treloar, ARCHER Project, Australian National Data Service,
Monash University, Australia
• Wanda Orlikowski, Information Technologies and Organization Studies ,
MIT Sloan School of Management
11
DATASPACE AGENDA - NSF SITE VISIT - Final - As of 7 Feb 2010 (v21)
Start Topic
Presenters
8:00 NSF Panel leaves hotel & gathers at MIT
8:30 1. INTRODUCTION & OVERVIEW (Vision & Rationale)
a. Stuart Madnick & MacKenzie Smith
8:50 2. SCIENTIFIC DOMAINS (Vision & Rationale)
Biological Oceanography
a. Ed DeLong
b. Terry Reese (Oregon State U)
Neuroscience / Neuroimaging
c. John Gabrieli
d. Steve White (Microsoft)
e. Tyler Walters (Georgia Tech)
f. Susan Hockfield (President, MIT) - arrives around 10am (approx)
10:15 Break
10:30 NSF Panel Closed Session #1
11:00 Additional Q&A
11:20 3. COMMUNITY BUILDING & PARTNERSHIPS (Activities, Organizational Structure)
Introduction
a. MacKenzie Smith
b. John Wilbanks (Science Commons)
Library community
c. Michele Kimpton (DuraSpace)
d. Geneva Henry (Rice)
e. Terry Reese (Oregon State U)
f. Tyler Walters (Georgia Tech)
Broader community
g. Stuart Madnick (for Google: Alon Halevy & MIST: Wei Lee Woon)
h. Joe Pato (for HP: Mei Hsu & Dejan Milojicic)
i. Stephen Todd (EMC)
j. Steve White (Microsoft)
k. Tom Malone (Citizen Science)
12:30
12:45
1:15
1:35
Citizen Science
Break
NSF Panel Closed Session #2 (Work ing lunch)
Additional Q&A
4. RESEARCH, DEVELOPMENT & OPERATIONS (Activities)
Development & Operations
Research agenda
a. MacKenzie Smith (Development, for Libraries, IS&T)
b. Marilyn Smith (Operations: IS&T)
c. Stuart Madnick (Research: data semantics, integration, quality,
etc)
d.
Hal Abelson (for DIG: Tim Berners-Lee, Danny Weitzner)
e. David Karger (Visualization)
Others
Break
NSF Panel Closed Session #3
Additional Q&A
5. SUSTAINABILITY & PROJECT MANAGEMENT
(Organizational Structure, DataNet Partner Leadership & Management)
a. MacKenzie Smith & Stuart Madnick
b. Michael Siegel
c. Ann Wolpert (Director, Libraries) & Claude Canizares (VP,
Research)
Others
5:30 6. WRAP-UP
a. Stuart Madnick & MacKenzie Smith
6:00 END
Minutes
Sub-totals
20
20
42
12
5
12
5
5
3
39
4
5
5
5
present
present
4
6
5
present
5
41
12
4
12
8
5
present
3:00
3:15
3:45
3:55
33
10
15
8
present
5
5
180
RE NSF Q&A TIME: It is assum ed that there w ill be som e brief speaker-specific Q&A after each speaker, then general Q&A for the rest of each segm ent.
12
Backup Slides
13
1. New types of science enabled
• Enhance scientific interdisciplinarity and innovation via
standards-based data architecture and broad adoption
• Disciplines: Neuroscience and Biological Oceanography
(a) Science and education goals help
– Library and Computer Science goals: minimize duplication of
effort, maximize access to prior work, improve interoperation
and quality
– Education goals: disseminated through multiple means (OCW)
to enable semantic tagging of data and reuse of data
(b) Metrics of Success
– Usage: number groups contributing and using, amount and
diversity of data shared and used, etc.
– Impact: Publications, discoveries
14
Neuroscience Domain
• Address questions, such as “Variation of
cognitive and emotions traits due to age?”
• Future requires access to large datasets, but
– Broadly distributed across many organizations
– Diverse types: DTI, fMRI, structural MRI, VBM
– Difficult to aggregate and annotate
• Initial organizations include
– Martinos Imaging Center (at MIT)
– Center for Advanced Brain Imaging (Georgia Tech)
– Collaboration with Microsoft
15
Biological Oceanography Domain
• Address questions such as “How does change in
ocean current cause proliferation of microbial
groups that, in turn, influence flux of carbon into and
out of the sea?”
• Need to interrelate diverse datasets
– Scale: from genome to biomes
– Types: 4D physical and biological oceanographic,
satellite, genomic, metagenomic, taxonomic,
nutrient analysis, bio-optical
– From diverse sources
• DataSpace will enable research not possible today
16
2. Value to Previous Investments
• For selected domains: Resources to organize, annotate,
archive, and publish existing data
–
–
–
–
Curated by partnership with library data curators
Improve collaborations, e.g., C-MORE (interrelate difficult)
Address complex legal, political, and social realities
Sustainability by providing significant new value to scientists
(e.g. ease of search, data integration, reuse)
(a) What data contributors gain from DataSpace
– More efficiently archive and reutilize their own data
– Able to utilize vast amounts of data from other sources
– Over time, will be respected academic achievement (citations)
(b) Investment utilized and enhanced
– Significant prior R&D by team members, e.g., Dspace, temporal
semantics, data quality and provenance, policy and legal, etc
17
3. Barriers to Implementation and Adoption
• In past, scientists often don’t participate because:
– Insufficient time and expertise (which we address via better
functionality and assistance from curators)
– Insufficient value back (which we address through re-use, etc)
• Some points:
– Demonstrable ease-of-use and value
• Especially sciences that are struggling with these problems
• Examples from Neuroscience and Biological Oceanography
–
–
–
–
–
Dedicated data curators
DataSpace Federation to represent collective needs
Openness: encourages scientific innovation and evolution
Support for policy and legal issues
Team has experience evolving systems (W3C, Dspace, etc.)
18
4. Cyberinfrastructure, Technical Sustainability
• Much of DataSpace cyberinfrastructure builds on prior
work (e.g. Dspace) and adds: (a) archive, (b) annotate to
enable discovery and re-use, (c) interoperate with Ed Tech,
“citizen science,” etc.
• Technical sustainability: Software free and open source –
establish architecture and standards
– Project will provide at least one reference implementation
– Enable multiple implementations (including commercial)
• Will develop cost and service models as exemplars
–
–
–
–
Institutions already expand large amounts
DataSpace will streamline, rationalize, distribute costs
Libraries have stood the test of time
Additional business models
• 1st Year Goal: Initial system and ingest of data, test interop
19
5. Manage Program, Providers, International
• Experience with highly distributed projects (Dspace)
• Management – see organization chart
– Multiple levels and multiple sub-groups
– Public/private partnership to insure industrial adoption and
relevance to other sectors
– Added management and data expertise from Advisory Board
• Data providers and assured participation
– Data initially from partners (Georgia Tech, MIT, OSU, Rice)
• Already communicating with scientists
– Then extend more broadly, initially to the DSpace community
• International Counterparts: (1) direct collaboration
(DuraSpace), (2) International partner (MIST), (3) International
corporations (EMC, Google, HP, Microsoft), (4) Advisory Board,
(5) indirect collaborations (C-MORE)
20
Download