ordbms1

Embed Size (px)

Citation preview

  • 8/4/2019 ordbms1

    1/18

    CERN/IT/DB

    Object / Object-Relational

    Database Management Systems

    Brief History & Outlook

    Jamie Shiers, IT-DB, CERN

  • 8/4/2019 ordbms1

    2/18

    CERN/IT/DB

    Overview Evolution of Object & Object-Relational DBMS

    Status of Market today

    Outlook

  • 8/4/2019 ordbms1

    3/18

    CERN/IT/DB

    DefinitionsFrom Stonebraker:Object-Relational DBMSs:The Next Great Wave

    Filesystems: no query

    RDBMS: query, simple data

    ODBMS: no (poor) query,complex data

    ORDBMS: query & complex (Picture Akmal Chaudri)

  • 8/4/2019 ordbms1

    4/18

    CERN/IT/DB

    RDBMS E.F.Codd: A Relational Model of Data for

    Large Shared Data Banks, CACM 13, 6

    (June 1970) pp 377 387 See http://wwwinfo.cern.ch/db/aboutdbs/papers/SRC-1997-018.pdf

    Led to numerous commercial products:Oracle, SQL/DS, DB-2, SYBASE, Informix,MS SQLServer, etc.

    In use at CERN since early 1980s

    Oracle Users Guide V2.3

    Relational S/W Inc.

  • 8/4/2019 ordbms1

    5/18

    CERN/IT/DB

    ODBMS Started ~late 1980s in research

    projects

    Altar p O2, Objectivity/DB, Ontos,

    Target markets (Cattell): CAD, CASE,CAP, Scientific/Medical, AI, etc.

    Object-Oriented DB Manifesto: An OO DBMS must satisfy 2 criteria:

    1. It should be a DBMS

    2. It should be object-oriented

  • 8/4/2019 ordbms1

    6/18

    CERN/IT/DB

    ODBMS Speculation

  • 8/4/2019 ordbms1

    7/18

    CERN/IT/DB

    ODBMS Reality Market much smaller than predicted

    Rationalisation

    but less than in RDBMS market O2 presumably now belongs to IBM!

    What happened?

  • 8/4/2019 ordbms1

    8/18

    CERN/IT/DB

    Market Numerous dinosaur analogies

    After all, IBM invented RDBMS technology and

    only now(?) have a competitive RDBMS product Can lightning strike twice / thrice?

    Predictions of ODBMS market: revised downin 1998 timeframe

    Original predictions (apparently) included ORDBMScomponent

    In other words, $1-2B includes Oracles share

  • 8/4/2019 ordbms1

    9/18

    CERN/IT/DB

    Technology ODBMS initially worked with Smalltalk,

    C++, CO2,

    Mainstream?

    Industry focus is now on Java, XML,

    The latter provided sufficient impetus

    for established vendors to enhance theirproducts

    See, for example, VLDB papers:

  • 8/4/2019 ordbms1

    10/18

    CERN/IT/DB

    VLDB Papers

    1989: complex objects / object modeling

    1990: performance of OODBs

    1999: XML; Extending SQL with user-defineddata types O-O, What Have They Done to DB2?

    2000: A Database Platform for Bioinformatics

    The need for complex data models has nowmoved mainstream

    The established vendors have seen it comingand have reacted

  • 8/4/2019 ordbms1

    11/18

    CERN/IT/DB

    Impact on ODBMS Down-turn of ODBMS market blamed

    on RDBMS announcements

    e.g. Oracle 8.0, 8i,

    Oracle 9i claims complete objectsupport according to SQL-99 standard

    C++ binding provides ODMG-likeinterface for C++ programmers withboth navigational & associative access

  • 8/4/2019 ordbms1

    12/18

    CERN/IT/DB

    RDBMS + objectextensions Can store ADTs

    Methods on server

    Complex Data withQueries

    $8B in 1996

    Likely to becomedominant DBMStechnology

    Complex Data

    Performance,

    scalability Tight Language

    Binding OQL - SQL3 query subset

    Growth similar toRDBMS in 80s

    ~$1B market by2001

    ODBMS / RDBMS / ORDBMS

    ~$100M?

    From 1999 RD45 Status report

  • 8/4/2019 ordbms1

    13/18

    CERN/IT/DB Risk Analysis:

    Summary of Options Evaluate C++ binding to e.g. ORACLE Add ESCROW clause to Objectivity contract

    Pursue possibility of source license

    Visit key Objectivity customers

    Produce new requirements list

    Estimate manpower to support Objy in house

    Estimate manpower for clean-sheet solution

    Continue to monitor alternatives

    The LCB agrees with the other suggested steps to mitigate risk,

    with the addition of trying to insure that user code in reconstruction

    and analysis programs is kept as standards compliant as possible.

  • 8/4/2019 ordbms1

    14/18

    CERN/IT/DB

    Espresso Preliminary results from the Espresso proof-of-concept

    prototype suggest that a home-grown ODBMS focussingon HEP needs could be built with a reasonable amountof manpower, perhaps some 15 person-years in total.However, it is far from clearwhere even such a modestamount of manpower could come from, particularly in a

    period when the overallCERN manpower is shrinking. Acertain amount of manpower could possibly be foundwithin the collaborations and at external institutes.However, a core team responsible for the maindevelopments would most likely have a significantimpact on the viability of such a project. In theimmediate future, there is essentially no manpoweravailable beyond that to complete the performance and

    scalability tests and write an architectural document. Inthe longer term, a home-grown ODBMS remains a

    possible fallback solution, although the time to developa full product suggests that this will no longer be truemuch later than 2002.

  • 8/4/2019 ordbms1

    15/18

    CERN/IT/DB

    Oracle C++ Binding Numerous C++ wrappers to Oracles C

    interface exist

    With Oracle 9i, a C++ binding,modelled on JDBC, comes with theproduct

    Had this existed in 1995, history mightwell have been different

  • 8/4/2019 ordbms1

    16/18

    CERN/IT/DB

    Oracle 9i studies Focus of this workshop Well, today & tomorrow

    Discussed at last Julys RD45workshopAs Oracle 8.2 the number changed but

    the features did not

    During the Hoffmann Review of LHCComputing

    Activity kicked off after IT Eloisemeeting

  • 8/4/2019 ordbms1

    17/18

    CERN/IT/DB

  • 8/4/2019 ordbms1

    18/18

    CERN/IT/DB

    Eloise ActionsThe possibility of a joint project betweenOracle and CERN should be explored to allowparticipation in the Oracle 9i F test with thegoals of evaluating this product as a potentialfallback solution and providing timelyfeedback on physics-style requirements. Non-

    staff human resources should be identifiedsuch that there is no impact on currentproduction services for Oracle andObjectivity.