johnz@latcs1.lat.oz.au (John Zeleznikow) (05/22/91)
DISTRIBUTED AND OBJECT - ORIENTED DATABASE SYSTEMS COURSE/WORKSHOP Lecturer: Dr Patrick Valduriez - Inria, Rocquencourt France Venue: La Trobe University Time: All day during week July 29 - August 2 1991 This course will involve designing and constructing both distributed and object-oriented database systems - technologies for the twenty first century. The course will consist of lectures, tutorials, laboratories and supervised practical work. Participants will have the opportunity to use the new object oriented database management system O2, developed at INRIA. Dr. Valduriez is currently the director of the Sabre Database Project at INRIA. He is the co-author of renowned books on distributed databases, and relational databases and knowledge bases. He gave a tutorial at VLDB90 on distributed and parallel databases. Dr. Valduriez will be supported by Dr. John Zeleznikow and his team at the Database Research Laboratory at La Trobe University. Dr. Zeleznikow is an associate editor of the Australian Computer Journal, Tutorial Chairman of VLDB90, and General Chairman of an IFIP WG2.6 Conference on Interoperable Databases. The fee for the course is $950 ($475 for academic staff and $95 for students and the unemployed), and includes all use of software and hardware, written materials and meals and refreshments. For further information and registration forms contact: Dr. John Zeleznikow Database Research Laboratory Applied Computing Research Institute La Trobe University Bundoora Victoria 3083 Australia Phone: 61.3.4791003 FAX: 61.3.4704915 ============================================================================ COURSE OUTLINE DISTRIBUTED AND OBJECT-ORIENTED DATABASE SYSTEMS Dr. Patrick Valduriez, INRIA, Le Chesnay, france DISTRIBUTED DATABASES CENTRALIZED VERSUS DISTRIBUTED DATA MANAGEMENT Critical features of relational databases: data independence, query optimization, reduction of data redundancy, transaction support. Additional objectives of distributed databases: distributed computing, site autonomy, increased reliability, extensibility and performance. Approaches to distributed data management. DISTRIBUTED DBMS ARCHITECTURES Transparencies: network, location,location, transparency. Architectural models: ISO/OSI, ANSI/SPARC, client-server model. Impact of standards: ISO SQL, IBM's SAA, ISO RDA, ISO TP. Global directory management issues. DESIGNING A DISTRIBUTED DATABASE Alternative design strategies: top-down vs. bottom-up design process. Distributed database design issues. Designing the best fragmentation: horizontal, vertical or hybrid. Selecting the best allocation of fragments to sites. PROCESSING AND OPTIMIZING QUERIES ON A DISTRIBUTED DATABASE Distributed query processing: problem, objectives, layers. Localization of distributed data: reduction rules for fragmentation. Optimization of distributed queries: cost model, database profiles, join ordering in fragment queries, algorithms. CONCURRENCY CONTROL AND RELIABILITY IN DISTRIBUTED DATABASE SYSTEMS Transaction management concepts: atomicity, consistency, isolation, durability. Distributed concurrency control: serializability theory, algorithms, distributed deadlock management. Distributed reliability: distributed protocols (two-phase commit, three- phase commit, network partitioning). INTEROPERABILITY USING DISTRIBUTED MULTIDATABASE SYSTEMS Integration of heterogeneous data models: schema integration. Multidatabase definition and manipulation within SQL. Multidatabase transaction management issues. MDBS developements: DB2 V2, INGRES/STAR, ORACLE/STAR, SABRINA*, SYBASE. IMPROVING PERFORMANCE AND RELIABILITY WITH PARALLEL DATABASE SERVERS Data Servers: objectives, data servers in distributed databases. Multiprocessor architectures: message-passing versus shared-memory. Parallel data placement: load balancing and replication issues. Parallel query processing. Parallel data servers: NONSTOP-SQL (Tandem), DBC (Teradata), BUBBA (MCC), GAMMA (U. of Wisconsin), GRACE (U. of Tokyo), EDS (ESPRIT). OBJECT-ORIENTED DATABASE SYSTEMS EVOLUTION OF DBMS REQUIREMENTS New application domains (OIS, CASE, CAD/CAM,AI, etc.). History of DBMS: theoretical and practical advances. Persitency and programming languages. Strenghts and weaknesses of Network and Relational DBMS. Objectives of OODBMS: richer data types, complex objects, computing power. THE OBJECT-ORIENTED APPROACH Objectives and history. OO concepts: object, class, method, message, inheritance, polymorphism. Advantages: encapsulation, modelling, modularity, code reusability. OO programming languages: C++, Smalltalk, Simula, Eiffel. Strenghts and weaknesses of OO. THE OODB APPROACH OODB concepts: persistence, sharing, identity, collections. Procedural vs. declarative programming: the impedance mismatch. OODB problems: persistence model, encapsulation, objects and values, class extensions, compilation and optimization, data control, schema evolution, transactions. OODB DATA MODELS Evolution of data models to capture more semantics. Extending the relational model with OO capabilities: INGRES OBJECT, ESQL. Extending an OO model with DB capabilities: ONTOS, GEMSTONE. Creating a new data model: ORION, O2. OODB LANGUAGES Extending SQL with abstract data types, object identity, complex objects: ESQL, OSQL. ISO extensions to SQL: SQL2, SQL3. Extending an OOPL: ONTOS with C++, OPAL with Smalltalk, O++. Extending a PL: ORION with LISP, O2-C and O2-C++. Query languages for OODB: the Object Management Group. IMPLEMENTING AN OODBMS Integration of OO and DB capabilities: difficulties. The three architectural approaches: extending an RDBMS: POSTGRES, SABRINA. extending an OO system: ONTOS, GEMSTONE, OBJECT-STORE, OBJECT-BASE. creating a new system: ORION, O2, GBASE, IRIS. Architecturing OODBMS in workstation/server environments. Implementation issues: object and memory management, versions, indexes, transactions, set-oriented operators. ANALYSIS AND COMPARISON OF OODBMS PRODUCTS Common description outline: objectives, model and language, architecture, implementation techniques, additional features. Analysis and comparison of POSTGRES, SABRINA, GEMSTONE, ONTOS, ORION, O2. TOWARDS HETEROGENEOUS, DISTRIBUTED OODB MANAGEMENT The toolkit approach to extensible object management: EXODUS (U. Wisconsin), GEODE (INRIA), ARJUNA (U. New Castle). Distributed object management (DOM): common object protocol, local application interfaces and distributed objects managers. Heterogeneous information systems integration: the DOM breadboard (GTE Labs.), the FUGUE project (Xerox). Issues: security, semantic integrity, physical integrity.