So Far (Schematically) yet So Near (Semantically)
0 likes442 views
The document discusses various types of incompatibilities encountered in enterprise integration, particularly focusing on schematic and entity definition heterogeneities. It outlines specific conflicts such as naming, data representation, scaling, precision, and integrity constraints, along with proposed solutions like creating virtual classes to standardize data across disparate databases. It highlights the importance of addressing these discrepancies for effective database interoperability.
So Far (Schematically) yet So Near (Semantically)
- 1. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Problems in Enterprise Integration: Schematic Heterogeneities between Semantically similar objects Based On: A. Sheth and V. Kashyap, So Far (Schematically) yet So Near (Semantically), Invited paper in Proceedings of the IFIP TC2/WG2.6 Conference on Semantics of Interoperable Database Systems, DS-5, November 1992 Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Schematic Heterogeneities Domain Definition Incompatibility Entity Definition Incompatibility Schematic Data Value Heterogeneity Incompatibility Abstraction Level Incompatibility Schematic Discrepancy
- 2. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Naming Homonyms Conflicts Data Representation Conflicts Synonyms Domain Data Scaling Incompatibility Conflicts Data Precision Conflicts Default Value Conflicts Attribute Integrity Constraint Conflicts Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Naming Conflicts • Semantically alike attributes may have different names and are called synonyms STUDENT(Id#, Name, Address) in database DB1 GRADUATE(SS#, Name, Address) in database DB2 STUDENT.Id# and GRADUATE.SS# are synonyms • Semantically unrelated attributes may have the same names and are called homonyms STUDENT(Id#, Name, Address) in database DB1 BOOK(Id#,Title, Author) in database DB2 STUDENT.Id# and BOOK.Id# are homonyms Solution: Synonyms can be renamed to the same concept. Homonyms can be renamed to different concepts. The associations are kept in a catalog
- 3. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Data Representation Conflicts Semantically alike attributes may have different data types or representations. STUDENT(Id#, Name, Address) in database DB1 GRADUATE(SS#, Name, Address) in database DB2 STUDENT.Id# is defined as INTEGER GRADUATE.SS# is defined as Char(9) Solution: The type of SS# is coerced to be an integer. A virtual class U_STUDENT is defined as follows: CREATE VCLASS U_STUDENT (Id: INTEGER, Name: Char(20), Address: Char(50)) AS SELECT Id#, Name, Address FROM DB1.STUDENT, SELECT SS# AS INTEGER, Name, Address FROM DB2.GRADUATE Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Data Scaling Conflicts Semantically alike attributes represented using different units and measures EMPLOYEE(Name, SS#, Salary) in database DB1 Salary is stored using British pounds EMPLOYEE(Name,SS#,Salary) in database DB2 Salary is stored using US dollars Solution: Use of appropriate conversion expressions. Assuming 1 British pound = 1.6 US dollars. A virtual class U_EMPLOYEE is defined: CREATE U_EMPLOYEE (Name: char(20), SS#: INTEGER, Salary: REAL) AS SELECT Name, SS#, 1.6*Salary FROM DB1.EMPLOYEE, SELECT Name, SS#, Salary FROM DB2.EMPLOYEE
- 4. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Data Precision Conflicts Semantically alike attributes represented using different precisions. Lower Upper Grade STUDENT(Name,SS#,Marks) in database DB1 81 100 A STUDENT(Name,SS#,Grade) in database DB2 61 80 B Grade is represented at a coarser level of precision when 41 60 C compared to Marks 21 40 D => Need to Map Marks to Grade represented in the Table 0 20 F Solution: Let the table MAP(Lower,Upper,Grade) be defined in DB1 A Virtual Class U_STUDENT can be defined: CREATE VCLASS U_STUDENT (Name: Char(20), SS#: INTEGER, Grade: CHAR(1)) AS SELECT Name, SS#, MAP.Grade FROM DB1.STUDENT, MAP WHERE Marks <= MAP.Upper and Marks >= MAP.Lower SELECT Name, SS#, Grade FROM DB2.STUDENT Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Default Value Conflicts Semantically alike attributes may have different default values in different databases. EMPLOYEE(Name,SS#,Salary,Bonus) in database DB1 where default(Bonus) = 10% EMPLOYEE(Name,SS#,Salary,Bonus) in database DB2 where default(Bonus) = 30% Solution: Default value for attribute of virtual class is used. Update of attribute value not propagated to databases. Suppose default(Bonus) for VCLASS U_EMPLOYEE is 20% Query retrieving values for bonus attribute will return a value corresponding to 20% However, update of Bonus value will not be propagated to DB1 or DB2
- 5. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Domain definition Incompatibilities Attribute Intergrity Constraint Conflicts Semantically alike attributes might be restricted by constraints that are not consistent with each other. Consider the attribute Age in the databases DB1 and DB2: C1: Age < 18 defined in database DB1 C2: Age > 21 defined in database DB2 C1 and C2 defined on attribute Age are inconsistent Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Database Identifier Conflicts Homonyms Naming Conflicts Entity Definition Incompatibility Union Compatibility Synonyms Conflicts Schema Isomorphism Conflicts Missing Data Item Conflicts
- 6. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Database Identifier Conflicts • Entity descriptions use identifier records that are semantically different • Also known as the key equivalence problem STUDENT(SS#, Course, Grades) in database DB1 STUDENT(Name, Course, Grades) in database DB2 SS# in DB1and Name in DB2 are semantically different identifiers Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Naming Conflicts • Semantically alike entities may have different names and are called synonyms EMPLOYEE(SS#, Name, Address) in database DB1 WORKER(SS#, Name, Address) in database DB2 EMPLOYEE and WORKER are synonyms • Semantically unrelated entities may have the same names and are called homonyms TICKET(Movie, Theater, Date, Timing) in database DB1 TICKET(Date, Offense, Officer, CourtDate) in database DB2 The entities TICKET in DB1 and DB2 are homonyms Solution: Synonyms can be renamed to the same concept. Homonyms can be renamed to different concepts. The associations are kept in a catalog
- 7. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Union Compatibility Conflicts Descriptors of semantically alike entities may not have equivalent signatures. STUDENT(SS#, Name, DateOfBirth) in database DB1 STUDENT(SS#, Name, Age) in database DB2 The attributes Age and DateOfBirth are not equivalent but can be made equivalent after some transformations CREATE VCLASS U_STUDENT (SS#: INTEGER, Name: Char(20), Age: INTEGER) AS SELECT SS#, Name, round(getdate()-DateofBirth) FROM DB1.STUDENT, SELECT SS#, Name, Age FROM DB2.STUDENT Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Missing Data Item Conflict Descriptors of semantically alike entities may have different number of attributes PROFESSOR(Name, OffPhone, HomePhone) in database DB1 PROFESSOR(Name, Phone) in database DB2 The attribute Phone in DB2 can be assumed to have both the Office and the Home Phone numbers CREATE VCLASS U_PROFESSOR (Name: Char(20), Phone: INTEGER) AS SELECT Name, OffPhone FROM DB1.PROFESSOR, SELECT Name, HomePhone FROM DB1.PROFESSOR, SELECT Name, Phone FROM DB2.PROFESSOR
- 8. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Entity definition Incompatibilities Missing Data Item with Implicit Value Descriptor of a similar entity may have a missing data item whose value is implicit STUDENT(SS#, Name, Address, Type) in database DB1 GRADUATE(SS#, Name, Address) in database DB2 STUDENT.Type can have the values “Grad” or “UnderGrad” GRADUATE has a missing data item whose value is implicit, i.e. GRADUATE.Type = “Grad” CREATE VCLASS U_STUDENT (SS#: INTEGER, Name: Char(20), Address: Char(50), Type: Char(9)) AS SELECT SS#, Name, Address, Type FROM DB1.STUDENT, SELECT SS#, Name, Address, Type FROM DB2.GRADUATE WHERE Type = “Grad” Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Data Value Incompatibilities Incompatibilities due to different values for the same entity Consider the entity SHIP(Id#,Name,Weight) SHIP(123,USSEnterprise,100) in database DB1 SHIP(123,USSEnterprise,200) in database DB2 Known Inconsistency Data Value Temporal Incompatibility Inconsistency Acceptable Inconsistency
- 9. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Data Value Incompatibilities Known Inconsistency When the source of Inconsistency is identified Solution: Choose the more reliable database In the previous example: If DB2 is identified as the more reliable database => weight(USSEnterprise) = 200 Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Data Value Incompatibilities Temporal Inconsistency • Inconsistency of a temporary nature. • Specified by using a temporal consistency predicate. In the previous example, the two weights may be considered temporarily inconsistent with DB1 updating it’s value subject to the following constraints: • At a particular date and/or at a specific point in time: AT 9:00 o’clock on 27th April 1995 • Before or after a specific instance of time or date: BEFORE 8:00 a.m. on 27th April 1995 • During intervals of time and/or date: INTERVAL(27th April 1995 AT 17:00 - 28th April 1995 AT 8:00) • Periodically, when certain amount of time has elapsed: Every Day AT 12:00
- 10. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Data Value Incompatibilities Acceptable Inconsistency Depending on the type of the query, the inconsistency between the values from different databases might be within an acceptable range. • Numerical Inconsistency: QUERY: Find the Tax Bracket of an Employee INCONSISTENCY: If abs(Salary1 - Salary2) < 0.99 The inconsistency is ignored • Non-numerical Inconsistency: QUERY: Find the State of residence of an Employee INCONSISTENCY: If Residence1 IN State1 and Residence2 IN State2 and State1 = State2 The inconsistency is ignored Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Abstraction Level Incompatibilities Generalization Conflicts Abstraction Level Incompatibility Aggregation Conflicts
- 11. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Abstraction Level Incompatibilities Generalization Conflicts When the same entity is represented at different levels of generalization STUDENT(SS#, Name, Major) in database DB1 GRAD-STUDENT(SS#, Name, Major, Advisor) in database DB2 A Virtual Class U_STUDENT is defined as follows: CREATE VCLASS U_STUDENT (SS#: INTEGER, Name: Char(20), Major: Char(20), Advisor: Char(20)) AS SELECT SS#, Name, Major, NA FROM DB1.STUDENT WHERE SS# NOT IN (SELECT SS# FROM DB2.GRAD-STUDENT), SELECT SS#, Name, Major, Advisor FROM DB2.GRAD-STUDENT Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Abstraction Level Incompatibilities Aggregation Conflicts When an entity in one database corresponds to a set of entities (or any other aggregation) in another CONVOY(Id#, AvgWeight) in database DB1 SHIP(Id#, InConvoy,Weight)in database DB2 A Virtual Class U_CONVOY is defined as follows: CREATE VCLASS U_CONVOY (Id#: INTEGER,AvgWeight: REAL) AS SELECT DB2.CONVOY.Id#, AvgWeight, ShipSet FROM DB1.CONVOY WHERE Exists (SELECT AvgWeight = avg(Weight) FROM DB2.SHIP WHERE SHIP.InConvoy = CONVOY.Id#)
- 12. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Schematic Discrepancies Data Value Attribute Conflict Schematic Attribute Entity Discrepancies Conflict Data Value Entity Conflict Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Schematic Discrepancies Data Value Attribute Conflict When attributes in one database corresponds to the value of the attributes in another Database DB1: relation STOCKS(Date, StkCode, ClsPrice) Database DB2: relation STOCKS(Date, Stk1, Stk2, ...) where Values(StkCode) = Stk1, Stk2, ..... and Values (Stk1, Stk2, ....) = (ClsPrice1, ClsPrice2, ....)
- 13. Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Schematic Discrepancies Attribute Entity Conflict When attributes in one database corresponds to entities in another Database DB2: relation STOCKS(Date, Stk1, Stk2) Database DB3: relations Stk1(Date, ClsPrice), Stk2(Date, ClsPrice) CREATE VCLASS U_STOCKS (Date: Char(7), StkCode: Char(4), ClsPrice: REAL) AS SELECT Date, “Stk1”, Stk1 FROM STOCKS, SELECT Date, “Stk2”, Stk2 FROM STOCKS, SELECT Date, “Stk1”, ClsPrice FROM Stk1, SELECT Date, “Stk2”, ClsPrice FROM Stk2 Template filename: ba-tmplt-vugrfDiffTitle.fm Version 1.0.3 Schematic Discrepancies Data Value Entity Conflict When entities in one database correspond to values of attributes in another Database DB1: relation STOCKS(Date, StkCode, ClsPrice) where values(StkCode) = Stk1, Stk2, .... Database DB2: relations Stk1(Date, ClsPrice) Stk2(Date, ClsPrice) ......