Involving New Scholars in Digital Libraries through the Networked Digital Library of Theses and Dissertations (NDLTD) Telcordia, Morristown, NJ December 30, 1999 Edward A. Fox fox@vt.edu CC CS DLRL Internet TIC Virginia Tech, Blacksburg, VA, USA Acknowledgements (Selected) Sponsors: ACM, Adobe, IBM, Microsoft, NSF, OCLC, US Dept. of Education, … Co-PIs: Marc Abrams, Robert Akscyn, John Carroll, John Eaton, Gail McMillan Students: Fernando Das Neves, Robert France, Marcos Goncalves, Neill Kipp, Paul Mather, Constantinos Phanouriou, James Powell, Ohm Sornil, David Watkins, Chang Zhang, Jianxin Zhao, … Information Systems at Virginia Tech – see online full version of short introduction to networking efforts, at http://rdweb.cns.vt.edu/talks/VT-Initiatives-9-99.ppt (by Jeff Crowder and Erv Blythe) OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion Virginia Tech Background Largest university in Virginia, land-grant, town population 35K plus 25K students, #2 in football Blacksburg Electronic Village, since 1992, with 80% of community on Internet Net.Work.Virginia, largest ATM network, with over 600 sites, for education, research, govt LMDS, Local Multipoint Distribution Service, gigabit wireless networking - 1/3 of Virginia Math Emporium, 500 workstations Faculty Development Initiative, round 2 Virginia Tech CS Department of CS focused on HCI since 1994 $2M (NSF RI) labs: usability, group decisions, info access Faculty (+ Kafura – OO/real-time, Head) – Abrams (Network Research Group, UIML – user interface) – Barfield (ISE - wearable) – Bowman (virtual environments and interface issues) – Carroll (design, scenarios, education, BEV) – Ehrich (equipment, graphics, BEV) – Hartson (theory & methodology, remote evaluation) – Hix (usability, VR/CAVE) – Ramakrishnan (data mining, recommemder systems) – Rosson (object orientation/languages, collaboration) – Shaffer (problem solving environments, education, GIS) – Williges (ISE - experimentation, meta-evaluation) ACITC Advanced Communications and Information Technology Center, opening summer 2000 Connects to the library, with a focus on IT 1/3 high-tech (multimedia) classrooms 1/3 digital/electronic library (reading room) 1/3 research labs: 10, including: – – – – Digital Library Research Laboratory (DLRL) Center for Applied Technologies in the Humanities HCI; HPC; Multimedia; Visualization (CAVE), … Spaces for industry-supported labs, visitors OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion Definition: Digital Libraries are complex systems that help satisfy info needs of users (societies) provide info services (scenarios) organize info in usable ways (structures) present info in usable ways (spaces) communicate info with users (streams) 5S Layers Societies Scenarios Spaces Structures Streams Definition: 5S Framework Societies: interacting people (, computers) Scenarios: services, functions, operations, methods Spaces: domains + constraints (e.g., distance, adjacency): 2D, vector, probability Structures: relations, trees, nodes and arcs Streams: sequences of items (text, audio, video, network traffic) (5 Element System: Fire, Wood, Earth, Metal, Water) 5S: Components Societies: roles, rituals, reasons, relationships, artifacts Scenarios: acquire, index, consult, administer, preserve Spaces: physical, temporal, functional, presentational, conceptual Structures: architectures, taxonomies, schema, grammars, links, objects Streams: granularities, protocols, paths, flows, turbulences Neill Kipp Dissertation Training interested groups about 5S and the Star Methodology, refining the Framework to have solid mathematical foundation Case studies of projects at Virginia Tech or involving VT staff/students: CSTC, NDLTD, NARA (National Archives, with SAIC), Lexis, ... Open also to study DL projects elsewhere Focusing too on the design artifacts developed and related issues of efficient description and representation (esp. with markup, hypermedia) Digital Libraries --- Virginia Tech MARIAN (NLM) CS DL Prototype - ENVISION (NSF, ACM) TULIP (Elsevier, OCLC) BEV History Base (NSF, Blacksburg) DL for CS Education - EI (NSF, ACM) WATERS, NCSTRL (NSF) NDLTD (SURA, US Dept. of Education) CSTC (NSF, ACM), CRIM (NSF, SIGMM) WCA (Log) Repository (W3C) VT-PetaPlex-1 (Knowledge Systems) Digital Library Courseware http://ei.cs.vt.edu/~dlib/ WWW pages or large PDF copy files Online quizzes based on book by Michael Lesk (Morgan Kaufmann Publishers) Contents based on book, with several other popular topics added (e.g., agents) Separate pages to supplement: Definitions, Resources (People, Projects), and References CS -> CSTC -> CRIM NSF and ACM Education Committee are funding a 2 year project “A Computer Science Teaching Center” CSTC - http://www.cstc.org/ College of NJ, U. Ill. Springfield, Virginia Tech Focus initially on labs, visualization, multimedia Multimedia part is also supported by a 2nd grant to Virginia Tech and The George Washington University: http://www.cstc.org/~crim/ (with curricular guidelines also under development) ACM will help provide reward for contributors, through Journal of Computing Education: Resources and Research (JoCERR?) Browsing (2) DLs Shorten the Chain from Editor Reviewer Publisher A&I Consolidator Library DLs Shorten the Chain to Author Teacher Digital Reader Editor Reviewer Learner Librarian Library Enhancing Learning with DLs Enhancing Learning Digital Libraries Adding to Digital Library (student) Interactive Experiences Using Digital Library (direct) (info literacy) Other Interactive Learning Activities Authoring (text, markup, hypermedia, cataloging-DC) Discovering, Browsing, Searching, Retrieving Indirectly Using Digital Library (embedded, by agent, ...) Submitting Work (ETD) (Metadata, PDF, XML) Annotating, Downloading, Installing, Feedback Using DL Contents (tools, data sets, env's, courseware, ...) Preserving (using stds, migrating, versioning) 5S Framework: Societies,Scenarios, Streams,Spaces, Structures Collaboration (in/around DL and its artifacts distance educ.) OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion A Digital Library Case Study Domain: graduate education, research Genre:ETDs=electronic theses & dissertations Submission: http://etd.vt.edu Collection: http://www.theses.org Project: Networked Digital Library of Theses & Dissertations (NDLTD) http:// www.ndltd.org ETDs Got Your Interest? ETD Web Site http://www.ndltd.org/ Graduate Students U. Laval Media Singapore AM Chronicle of Higher Ed. National Public Radio NY Times ... Key Ideas: Scalability Networked infrastructure University collaboration Workflow, automation Maximal access Education is the rationale 8th graders vs. grads Authors must submit Standards PDF, SGML, MM MARC, DC, URNs Federated search What are we doing? Aiding universities to enhance grad educ., publishing and IPR efforts Helping improve the availability and content of theses and dissertations Educating ALL future scholars so they can publish electronically and effectively use digital libraries (i.e., are Information Literate and can be more expressive) What are the long term goals? 400K US students / year getting grad degrees are exposed / involved 200K/yr rich hypermedia ETDs that may turn into electronic portfolios Dramatic increase in knowledge sharing: lit. reviews, bibliographies, … Services providing lifelong access for students: browse, search, prior searches, citation links What led to today’s meeting? 1987 mtg in Ann Arbor: UMI, VT, … 1992 mtg in Washington: CNI, CGS, UMI, VT and 10 universities with 3 reps each 1993 mtg in Atlanta to start Monticello Electronic Library (MEL): SURA, SOLINET 1994 mtg in Blacksburg re ETD project: std of PDF + SGML + multimedia objects 1996 funding by SURA, US Dept. of Education (FIPSE) for regional, national projects 1997 meetings in UK, Germany, ... Sept. 1999 meeting in Paris at UNESCO Headquart. http://www.unesco.org/webworld/etd/ Status of the Local Project Approved by university governance Spring 1996; required starting 1/1/97 Submission & access software in place Submission workshops for students (and faculty) occur often: beginner/adv., focused on Adobe PDF and multimedia formats Faculty training as part of Faculty Development Initiative Over 2000 ETDs in collection Library Costs $12/vol. for paper thesis processing – catalog, bind, security strip, label, and shelving – @950 vols./yr. = $11.4K $3.20/vol. ETD processing – cataloging @950 vols./yr. = $3040 $.07/vol. Shelving (save 166 ft/yr) $.04/vol. Circulation (of 3000 copies/yr) Student Prepares Thesis or Dissertation NDLTD Literature Computer Resources Research Student Defends and Finalizes ETD My Thesis ETD Student Gets Committee Signatures and Submits ETD Signed Grad School Graduate School Approves ETD Student is Graduated Ph.D. Library Catalogs ETD and New Students Have Access to the New Research WWW NDLTD OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion Institutional Members Coalition for Networked Information (CNI) Committee on Institutional Cooperation (CIC) Diplomica.com Dissertation.com Dissertationen Online (Germany) Ibero-American Science & Technology Education Consortium (ISTEC, www.istec.org) National Library of Portugal (for all universities) Organization of American States (SEDI/OAS) UNESCO (www.unesco.org/webworld/etd) US University Members (35+) Air University (Alabama) Brigham Young University Cal Tech Clemson University College of William & Mary Concordia University (Illinois) East Carolina University East Tenn. State University Florida Institute of Tech. Florida International University George Washington University Marshall University (W. Va.) Miami U. of Ohio MIT (in process) Michigan Tech Naval Postgraduate School (CA) North Carolina State U. Penn. State University Rochester Institute of Tech. U. of Colorado Health Sci. Cntr. U. of Florida U. of Georgia University of Hawaii, Manoa U. of Iowa U. of Maine U. of Oklahoma U. of South Florida U. of Tennessee, Knoxville U. of Tennessee, Memphis U. of Texas at Austin U. of Virginia U. Wisconsin - Madison Vanderbilt U. Virginia Tech - required since 1/97 West Virginia U. - required fall 1998 Worcester Polytechnic Inst. Australian Project Members U. New South Wales (lead institution) U. of Melbourne U. of Queensland U. of Sydney Australian National University Curtin U. of Technology Griffith U. German Project Members Humboldt University (lead institution) 3 other universities 5 learned societies – Mathematics, Physics, Chemistry, Sociology, Education 1 computing center 2 major libraries CBUC (www.cbuc.es, Spain) Consorci de Biblioteques Universitàries de Catalunya, as group, with 9 members: – – – – – – – – – Universitat de Barcelona Universitat Autonòma de Barcelona Universitat Politècnica de Catalunya Universitat Pompeu Fabra Universitat de Girona Universitat de Lleida Universitat Rovira i Virgili Universitat Oberta de Catalunya Biblioteca de Catalunya Other International Members Chinese University of Hong Kong Chungnam National U., Dept of CS (S. Korea) City University, London (UK) Darmstadt U. of Tech. (Germany) Free University of Berlin (Germany - Vet. Med.) Gyeongsang National U. (Korea) India Institute of Technology, Bombay (India) Nanyang Technological U. (Singapore, part) National U. of Singapore (Singapore, part) Polytechnic University of Valencia (Spain) Rhodes U. (South Africa) St. Petersburg St. Tech.U (Russia) Univ. de las Américas Puebla (Mexico) Univ. of Alicante (Spain) Univ. of Pisa (Italy) U. Laval; U. of Guelph; U. Waterloo; Wilfrid Laurier U. (Canada) 0 Date Joined 11/11/99 9/11/99 7/11/99 5/11/99 3/11/99 1/11/99 11/11/98 9/11/98 7/11/98 5/11/98 3/11/98 1/11/98 11/11/97 9/11/97 7/11/97 5/11/97 3/11/97 Number of Members NDLTD Members 80 70 60 50 40 30 20 10 Popular Works 1996 458 Seevers, Gary L. Identification of Criteria for Delivery of Theological Education Through Distance Education: An International Delphi Study (Ph.D., Educational Research and Evaluation, April 1993; 1353Kb) 432 Hohauser, Robyn Lisa. The Social Construction of Technology: The Case of LSD (MS in Science and Technology Studies, Feb. 1995; 244Kb) 390 Childress, Vincent William. The Effects of Technology Education, Science, and Mathematics Integration Upon Eighth Grader's Technological Problem-Solving Ability (Ph.D. in Vocational and Technical Education, July 1994; 285Kb) 310 Kuhn, William B. Design of Integrated, Low Power, Radio Receivers in BiCMOS Technologies (Ph.D. in Electrical Engineering, Dec. 1995; 2Mb) 287 Sprague, Milo D. A High Performance DSP Based System Architecture for Motor Drive Control ( MS in Electrical Engineering, May 1993; 878Kb) 165 Wallace, Richard A. Regional Differences in the Treatment of Karl Marx by the Founders of American Academic Sociology (MS in Sociology, Nov. 1993; 479Kb) 150 McKeel, Scott Andrew. Numerical Simulation of the Transition Region in Hypersonic Flow (Ph.D. in Aerospace Engineering, Feb. 1996; 3Mb) Popular Works 1997 9920 Liu, Xiangdong. Analysis and Reduction of Moire Patterns in Scanned Halftone Pictures (Ph.D. in Computer Science, May 1996; 6.6Mb) 7656 Petrus, Paul. Novel Adaptive Array Algorithms and Their Impact on Cellular System Capacity (Ph.D. in Electrical Engineering, March 1997; 5Mb) 2781 Agnes, Gregory Stephen. Performance of Nonlinear Mechanical, Resonant-Shunted Piezoelectric, and Electronic Vibration Absorbers for Multi-Degree-of-Freedom Structures (Ph.D. in Engineering Mechanics, Sept. 1997; ? + 7926Kb) 2492 Gonzalez, Reinaldo J. Raman, Infrared, X-ray, and EELS Studies of Nanophase Titania (Ph.D. in Physics, July 1996; 4607Kb) 1877 Shih, Po-Jen. On-Line Consolidation of Thermoplastic Composites (Ph.D. in Engineering Mechanics, Feb. 1997; 3.3Mb) 1791 Saldanha, Kevin J. Performance Evaluation of DECT in Different Radio Environments (MS in Electrical Engineering, Aug. 1996; 3.2Mb) 1431 DeVaux, David. A Tutorial on Authorware (MS in CS, April 1996; 2.3Mb) 1394 Kuhn, William B. Design of Integrated, Low Power, Radio Receivers in BiCMOS Technologies (Ph.D. in Electrical Engineering, Dec. 1995; 2518Kb) Usage of ETDs in VT Collections Total requests Daily Requests Abstract requests Hosts served 1996 1997 1998 37,171 247,537 465,974 1999 Jan-Aug 907,104 102 685 1,722 3,121 25,829 112,633 177,647 143,056 9,015 22,725 28,022 52,663 International Use 1996 850 608 346 713 387 463 250 191 183 22 83 1997 2992 2,501 2378 2367 1264 1161 725 867 1130 967 958 1998 8170 United Kingdom 4223 Australia 7373 Germany 3970 Canada 2201 South Korea 4431 France 2553 Italy 2781 Netherlands 1449 Brazil 1089 Thailand 1414 Greece OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion Relationship with publishers Concern of faculty and students that still wish to publish books or journal articles, voiced: campus, Chronicle, NPR, Times Solution: Approval Form gives students, faculty choices on access, when to change access condition; use IPR controls in DL Solution: by case, work with publishers and publisher associations to increase access – – AAP, AAUP AAAS, ACM, ACS, Elsevier, ... Some responses from publishers ACM: need to acknowledge copyright Elsevier: need to acknowledge copyright IEEE-CS: endorse initiative ACS: After first publication, can release Textbook publishers: different market, manuscript significantly reworked General: restricting access to local campus will not cause any problems Survey by Joan Dalton, Canada For professional societies Like “writing across the curriculum” Besides writing: computing/communications, information literacy, personal digital library management, tool use, research methods, collaboration, archiving/preservation Data sets, communities of users of them Classification systems / browsing / searching National Research Council (NRC) booklet “On becoming a researcher in the digital age” Who are sponsors / cooperators? Funding, Donations of hardware/software – – – – – – SURA US Dept. of Education (FIPSE) Adobe Systems IBM Microsoft OCLC Others Serving on Steering Committee – National/Regional Projects: Australia, French speaking group, Germany, IberoAmerica (ISTEC), UK (UTOG) – Council of Graduate Schools, National Lib. Canada, NSF, OAS, SOLINET, UMI, UNESCO, ... How does this relate to UMI? (Bell and Howell) 1987 UMI workshop to explore ETDs Support letter for US Dept. of Ed. proposal Steering and technical committee membership ProQuest Direct pilot of scanning works started 1/1/97, free 2 yr access to front part Collaborating on: – – accepting electronic author submissions standards (e.g., representation), research ETD Initiative (and UMI) Students Learn about DL, EPub TDs become more expressive Global TDs become more accessible, archived Universities UMI N. Amer. (T)Ds are accessible, archived How can a university get involved? Select – – – – planning/implementation team Graduate School Library Computing / Information Technology Institutional Research / Educ. Tech. Send us letter, give us contact names Adapt Virginia Tech solution – – Build interest and consensus Start trial / allow optional submission Convene Local Planning Group ETD Build Local ETD Site ETD Workshop/Training Digital Library Policies Inspection/Approval Type 1 Members University Requires ETDs Adobe Acrobat and/or XML/SGML tools Automated submission & processing Archive/access - UMI, (OCLC, Center for Research Libraries, Virginia Tech,) ... (Local) WWW site, publicity (Local) Assistance provided as requested: email, phone, listserv(s) Type 2 Members University Agrees to Require ETDs Like Type 1 but set date not reached Usually has an option or pilot May: wait for new AY; start with all who enter after; … Build grass roots support – – – – Advisory committee: representative? expert? Champions to spread by word of mouth Approval: Senates, Commissions, Deans, Students Publicity to reach community NDLTD Members, Types 3-7 3. Part of university requires ETDs 4. University allows ETDs 5. University investigating, has pilot 6. University consortium joins: – CBUC (Catalunya, Spain) 7. Non-university organization joins – CNI (Coalition for Networked Info.) – ISTEC, OAS, UNESCO, … Everyone Learns Students become “info literate” Students learn about discovery, search, categorization/classification (e.g., CoRC), e-pub (e.g., XML, multimedia, hypertext), preservation, helping others find/reuse Campus starts to think about IPR – e.g., Virginia Tech symposium http://www.rgs.vt.edu/resmag/seminars.html Faculty and students improve quality as reader base expands OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion User Search Support (multilingual, XML) NDLTD World Federated Search User Interface Virginia Tech ... (univ) UMI ... (corporate) OhioLink ... (univ group) Portugese NL ... (national lib) Australia (regional) Note: All groups shown are connected with NDLTD. www.theses.org James Powell student project, D-Lib Magazine description in Sept. 1998 XML description of each site – type of search engine / service – language – coverage (for resource discovery) Adding Z39.50 gateway capability and integrating with MARIAN, along with Harvest and Open Archives protocols (according to Santa Fe Convention) – see www.openarchives.org Open Archives Initiative Santa Fe meeting, Oct. 21-22, 1999 Workshop early June, San Antonio, DL’00 LANL, CNI, DLF, Mellon, … Convention Archives -> Open Archives – – – – Support unique archive identifiers Implement Open Archives Metadata Set (DC-based, using XML) Implement Dienst harvesting interface Register the archive Build tools, layer other services: linking, searching, … Approaches to Open Archives Build By institution Build By discipline Author Category Interdisciplinary Year Language Query … Open Archives Members Original Participants in the Open Archives Initiative – – – – – – – – – – – – – – – – – – – – – Caroline Arms, Library of Congress Leslie Carr, University of Southampton Mark Doyle, American Physical Society Dale Flecker, Harvard University Edward A. Fox, Virginia Tech Michael Friedman, HighWire Press, Stanford University Paul M. Gherman, Vanderbilt University Paul Ginsparg, Los Alamos National Laboratory & xxx Stevan Harnad, University of Southampton Thomas Krichel, University of Surrey & RePEc Carl Lagoze, Cornell University Rick Luce, Los Alamos National Laboratory Clifford Lynch, Coalition for Networked Information Kurt Maly, Old Dominion University Michael L. Nelson, NASA Langley Research Center John Ober, California Digital Library Bob Parks, Washington University & EconWPA Herbert Van de Sompel, University of Ghent Eric F. Van de Velde, California Institute of Technology Don Waters, The Andrew W. Mellon Foundation Ken Weiss, California Digital Library Others Joining (selected) – – – – – University of Virginia – Jim French, Worthy Martin, Thornton Staples, NEC Research Institute - C. Lee Giles and Steve Lawrence Internet Archive - Kurt Bollacker, Marlita Kahn India - University of Mysore – Shalini Urs Mexico – University of Monterrey - David Garza Salazar Access Approaches Goal: Maximize access and services, e.g., by encouraging: UMI centralized services Distributed service: Dienst, Z39.50, … Regional services (e.g., OhioLink) Global service: Open Archives Local servers with browse, search From local catalogs to local archives WWW robot indexing and search services – Access Possibilities Web search engines www. theses. org Virginia MIT National Tech Library of Portugal www. library openarchives. catalog org clients CBUC (Spain) Ohio Link 3rd Party Services (e.g., Bell & Howell) National Projects: AU, GE, … Support Services Developed WWW site with > 300 Mb, CD, videotape Automated submission system (MySQL, UNIX, WWW scripts - grad school/library) Student guidelines, style sheets, multimedia training materials, FAQs, press info SGML and XML DTDs for ETDs SGML to HTML (web generator) LaTeX, Word templates, converters Support Offered Software, documentation, tech support Email, listservs (etd-l@listserv.vt.edu, eval, -grad, -library, -technical) Donations: Adobe, Microsoft Evaluation: instruments, analysis http://scholar.lib.vt.edu - solutions/statistics (Temporary storage / archiving; aid - in setting up an int’l service & archive) Enhancements Dublin Core spec, MARC crosswalk DTDs for SGML, XML(+ <discipline>ML) Annotation system (author, friends, notes) Routing system (based on Sift) Multilingual WWW site, training materials Better federated search (w. Z39.50, planned with Dienst and Harvest - using MARIAN) Integrate SFX, CiteSeer (citation database and linking, plagiarism detection) Accessibility Activities / Plans Interface design (simple, 3D, VR) Usability studies Generic multi-lingual support Support for those with disabilities Hybrid collection (paper, MARC, abstracts, full-text, multimedia) Disciplinary classifications, tools Visualization of results, collection SPIRE Visualization CAVE Experiments Use a familiar metaphor – building / floor / room / shelf / book Rearrange orderings / shelving – use categories, clustering, ranking – use visualization: colors and gaps – study space mappings: physical, logical Simplify movement for key tasks ENVISION NSF “A User-Centered Database from the Computer Science Literature” (1991-93) Collected bib/typesetter data, converted to SGML Scanned thousands of page images MARIAN search engine - can be made available (also applied to the Virginia Tech library catalog) used as part of a prototype object-based DL, with tailored visualization interface (L. Nowell dissertation) Envision Results Window MARIAN Multiple Access Retrieval of Information with ANnotations (Musical: Marian the Librarian …) Evolved from 1980’s CODER system to a distributed Online Public Access Catalog (OPAC), then DL backend, now becoming a full DL system From C/C++ to Java by Jianxin Zhao Future uses: NDLTD, NUDL, PetaPlex MARIAN Layers User User User User Interface Layer User Information Layer Search Engine Layer Database Layer User MARIAN Testing Architecture Load Generator Webgate Java Server C/C++ Server MARIAN Parallelism response time (ms) Java part response time vs. query rate comparation (type 1 requests) 4000 3000 2000 1000 0 0 100 200 300 queryrate (#/min) all modules in one machine one "webgate" two "webgate"s four "webgate"s 400 500 MARIAN Response Time time delay (ms) Four "webgate"s, decomposed time delay vs. query rate 4000 3000 2000 1000 0 0 100 200 300 query rate (#/min) system after Java server 400 500 France Dissertation Key developer since CODER Applying computational linguistics efforts with machine readable dictionaries Applying opportunistic handling of term lists for ranking, usable displays (“to be or not to be, that is the”) Developing and evaluating variety of interfaces PetaPlex Digital Library Machine (“super” object store) Parallel computer / storage utility for scale of 1000 to 100,000,000 gigabytes (1 Tbyte - 100 Pbyte) Knowledge Systems Incorporated supplied VTPetaPlex-1 for $250,000 with – high speed backbone connections (eventually 1 Gbps) – 2.5 terabytes through 100 “Nanoservers”: – Each = Network connection + IBM 25GB disk + 233 MHz Pentium II + Linux Service Machine 1 Service PetaPlex Complex Service Machine 2 Nanoserver FRONT END MACHINE RS/6000, 1G RAM, 4 Proc. Machine 3 Service Machine 4 Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver Nanoserver PetaPlex Service Machines Front-end provides handle/repository abstraction through hashing Small object server Large object server – video on demand – streaming audio Information retrieval server Proxy / cache server (e.g., 1 terabyte server of 1000 worldwide for Comsat/Intelsat) PetaPlex Top View 4 ft. side PetaPlex Side View 15 Roles: * Support * Cooling * Power shelves 8 ft. high 4 ft. wide PetaPlex Cost Goals, Approach Maximize number of seeks achievable Maximize % of cost invested in disks Maximize flexibility and reliability Minimize cost per unit of storage Approach “information utility” Increase throughput and reliability by replicating on other PetaPlex systems Use robotics, wireless, and commodity production of nanoservers Sornil & Mather Dissertations Proposing 100 Tbyte wireless Petaplex for $2M Mather: efficiently handling very large numbers of objects of varying sizes Sornil: efficiently handling IR for very large dynamic collections, large numbers of users, high transaction rates, large inverted files – modeling and simulation – data organization – parallelization of algorithms, alone and in combination for retrieval (related) tasks Problems Addressed THE PETAPLEX-TYPE ARCHITECTURE Early Results PARTITIONING SCHEMES: A PRELIMINARY ANALYSIS To preliminarily study effects of three important parameters on the performance of the partitioning schemes Term selection characteristic Number of queries in the system Number of disk nodes To have a preliminary performance trend of the Hybrid partitioning scheme Inverted File Partitioning Schemes: Term Partitioning Document Partitioning Hybrid Partitioning Early Results Effects of Number of Disk Nodes Skewed Term Selection (with 1024 queries in the system) Uniform Term Selection (with 1024 queries in the system) OUTLINE Introduction Digital libraries NDLTD case study Members, statistics Relationships, universities Access, software, hardware Conclusion DL Challenges Preservation - so people with trust DLs infrastructure – affordable storage in large capacity, very fast networks, ... Supporting Scalability, sustainability, interoperability DL industry - critical mass by covering libraries, archives, museums, corporate info, govt info, personal info - “quality WWW” integrating IR, HT, MM, ... – Need tools & methods to make them easier to build