Transcript Presentation

Automatic Conflict
Resolution to Integrate
Relational Databases
Ramon Lawrence
University of Manitoba
[email protected]
Outline
 Introduction,
Motivation, and Background
 The integration architecture
 Standard
 Example
dictionary, X-Specs, query processor
integration
 Northwind,
 Querying
Southstorm databases
the integrated databases
 Generating
SQL queries from semantic queries
 Unity
implementation
 Contributions, Conclusions, and Future Work
Page 2
What is Integration?
 Two
levels of integration:
 Schema
integration - the description of the data
 Data integration - the individual data instances
 Integration
problems include:
 Different
data models and conflicts within a model
 Incompatible concept representations
 Different user or view perspectives
 Naming conflicts (homonym, synonym)
 Integration
handles the different mechanisms for
storing data (structural conflicts), for referencing
data (naming conflicts), and for attributing
meaning to the data (semantic conflicts).
Page 3
Why is Integration Required?
 There
are many integration environments:
 Operational
systems within an organization
 System integration during company merger
 Data warehouses, Intranets, and the WWW
 Users
require information from many data
sources which often do not work together.
 Companies require a global view of their entire
operations which may be present in numerous
operational databases for different departments
and distributed geographically.
 E-commerce demands integration of web
databases with production systems.
Page 4
Previous Work
 Research
systems:
 integrating
systems by logical rules (Sheth)
 defining global dictionaries (Castano)
 Carnot Project using the Cyc knowledge base
 wrapper and mediator systems:
 Information Manifold, TSIMMIS, Infomaster
 Industrial
systems and standards:
 Metadata
Interchange Specification (MDIS)
 XML, BizTalk, E-commerce portals
 Query
Languages:
 SQL,
MSQL, IDL, DIRECT, SchemaSQL
Page 5
Previous Work Summary
 Current
techniques for database integration have
some of these problems:
 Require
integrator to understand all databases
 Integration process is manual
 Do not hide system complexity from the user
 Force changes on the existing database systems
 Construct global view manually
 Suffer from query imprecision (query containment)
Page 6
Our Approach
 Our
approach combines standardization and
query mapping algorithms.
 The major idea is that schema conflicts can be
resolved if we:
 Eliminate
all naming conflicts
 Define a language capable of determining schema
equivalence and performing transformations
 Naming
conflicts are eliminated by accepting a
standard term dictionary.
 Not
a knowledge base or set of mediated views
 Leverages semantic information in English words
Page 7
Integration Architecture
Client
Client
Multidatabase Layer
• user’s view of integration
2) X-Spec Editor
Integrated Context View
X-Spec
Editor
Standard
Dictionary
Architecture Components:
1) Integrated Context View
• stores schema & metadata
• uses XML
Integration
Algorithm
3) Standard Dictionary
• terms to express semantics
4) Integration Algorithm
Query Processor and ODBC Manager
5) Query Processor
Subtransactions
X-Spec
X-Spec
Database
Database
Local Transactions
• combines X-Specs into
integrated context view
• accepts query on view
• determines data source
mappings and joins
• executes queries and
formats results
Architecture Components
 The
architecture consists of four components:
A
standard dictionary (SD) to capture data semantics
 SD terms are used to build semantic names describing
semantics of schema elements.
 X-Specs
for storing data semantics
 Database metadata and semantic names stored using XML
 Integration
Algorithm
 Matches concepts in different databases by semantic names.
 Produces an integrated view of all database concepts.
 Query
Processor
 Allows the user to formulate queries on the view.
 Translates from semantic names in integrated view to SQL
queries and integrates and formats results.

Involves determining correct field and table mappings and
discovery of join conditions and join paths
Page 9
Integration Processes
 The
integration architecture consists of three
separate processes:
 Capture
process: independently extracts database
schema information and metadata into a XML
document called a X-Spec.
 Integration
process: combines X-Specs into a
structurally-neutral hierarchy of database concepts
called an integrated context view.
 Query
process: allows the user to formulate queries on
the integrated view that are mapped by the query
processor to structural queries (SQL) and the results
are integrated and formatted.
Page 10
Integration Architecture:
The Capture Process
Relational
Schema
Automatic
Extraction
Specification
Editor
Standard
Dictionary
 Capture
X-Spec
DBA Lookup
of terms
process involves:
 Automatically
extracting the schema information and
metadata using a specification editor
 Assigning semantic names to each schema element
(tables and fields) to capture their semantics
Page 11
Architecture Components:
The Standard Dictionary
A
standard dictionary (SD) provides standardized
terms to capture data semantics.
 Hierarchy
of terms related by IS-A or Has-A links
 Contains base set of common database concepts, but
new concepts can be added
A
SD term is a single, unambiguous semantic
definition.
 Several
SD entries for a single English word are
required if the word has multiple definitions.
 The
top-level dictionary terms are those
proposed by Sowa.
Page 12
Architecture Components:
Dictionary vs. Knowledge Base
 The
standard dictionary differs from a knowledge
base such as Cyc because:
 Not
intended to be a general English dictionary or
contain knowledge facts about the world
 Dictionary is evolved as new terms are required
 Not all English words are used
 Dictionary
provides the systems with no “knowledge”
 Since no facts are stored, system cannot deduce new facts
 Dictionary terms are just semantic place holders, integrators
determine the semantics of the database not the system
 Simplified
organization
 Dictionary is organized as a tree for efficiency and simplicity
in determining related concepts
 Re-use
of terms
 Terms are re-used in semantic names
Page 13
Architecture Components:
Using the Standard Dictionary
 SD
terms are used to build semantic names
describing semantics of schema elements.
 Semantic names have the form:
 semantic
name := [CT_Type] | [CT_Type] CN
 CT_Type := CT | CT {; CT} | CT {,CT}
 CT := context term, CN := concept name
 each CT and CN is a single term from the SD
 Semantic
names are included in specifications
describing a database.
Page 14
Northwind & Southstorm
Integration Example
Northwind Database Schema
Tables
Categories
Customers
Employees
OrderDetails
Order
Products
Shippers
Suppliers
Fields
CategoryID, CategoryName
CustomerID, CompanyName
EmployeeID, LastName, FirstName
OrderID, ProductID, UnitPrice, Quantity
OrderID, CustomerID, EmployeeID, OrderDate,
Shipvia
ProductID, ProductName, SupplierID, CategoryID
ShipperID, CompanyName
SupplierID, CompanyName
Page 15
Northwind & Southstorm
Integration Example (2)
Southstorm Database Schema
Tables
Fields
Orders_tb Order_num, Cust_name, Emp_name, Item1_id, Item1_qty,
Item1_price, Item2_id, Item2_qty, Item2_price
Page 16
Integration Example (3)
Northwind Semantic Name Mappings
Type
T
F
F
T
F
F
T
F
F
F
T
F
F
F
F
Semantic Name
[Category]
[Category] Id
[Category] Name
[Customer]
[Customer] Id
[Customer] Name
[Employee]
[Employee] Id
[Employee] Last Name
[Employee] First Name
[Order;Product]
[Order] Id
[Order;Product] Id
[Order;Product] Price
[Order;Product]
Quantity
System Name
Categories
CategoryID
CategoryName
Customers
CustomerID
CompanyName
Employees
EmployeeID
LastName
FirstName
OrderDetails
OrderID
ProductID
UnitPrice
Quantity
Type
T
F
F
F
F
F
T
F
F
F
F
T
F
F
T
F
F
Semantic Name
[Order]
[Order] Id
[Order;Customer] Id
[Order;Employee] Id
[Order] Date
[Order;Shipper] Id
[Product]
[Product] Id
[Product] Name
[Product;Supplier] Id
[Product;Category] Id
[Shipper]
[Shipper] Id
[Shipper] Name
[Supplier]
System Name
Orders
OrderID
CustomerID
EmployeeID
OrderDate
Shipvia
Products
ProductID
ProductName
SupplierID
CategoryID
Shippers
ShipperID
ShipperName
Suppliers
[Supplier] Id
[Supplier] Name
SupplierID
SupplierName
Page 17
Northwind & Southstorm
Integration Example (4)
Southstorm Semantic Name Mappings
Type
Table
Field
Field
Table
Field
Field
Table
Field
Field
Field
Semantic Name
[Order]
[Order] Id
[Order;Customer] Name
[Order;Employee] Name
[Order;Product] Id
[Order;Product] Quantity
[Order;Product] Price
[Order;Product] Id
[Order;Product] Quantity
[Order;Product] Price
System Name
Orders_tb
Order_num
Cust_name
Emp_name
Item1_id
Item1_qty
Item1_price
Item2_id
Item2_qty
Item2_price
Page 18
What is a semantic name?
A
semantic name is a universal, semantic
identifier in a domain.
 Similar
to a field name in the Universal Relation.
 Semantics are guaranteed unique by construction.
 System has mechanism for comparing semantics
across domains even though it does not understand
them. (Exploiting semantics in English words.)
 Important
 context
definitions:
- a semantic name is a context if it maps to a table
 concept - a semantic name is a concept if it maps to a field
 context closure - of semantic name Si denoted Si* is the
set of semantic names produced by taking ordered subsets
of the terms of Si = {T1, T2 , … TN} starting with T1.
Page 19
Architecture Components:
X-Specs
 Database
metadata and semantic names are
combined into specifications called X-Specs:
 Stored
and transmitted using XML
 Contains information on a relational schema
 Organized into database, table, and field levels
 Stores semantic names to describe and integrate
schema elements
Page 20
Southstorm X-Spec
<?xml version="1.0" ?>
<Schema name = "Southstorm_xspec.xml” xmlns="urn:schemas-microsoft-com:xml-data"
xmlns:dt="urn:schemas-microsoft-com:datatypes">
<ElementType name="[Order]" sys_name = "Orders_tb" sys_type="Table">
<element type = "[Order] Id" sys_name = "Order_num" sys_type = "Field"/>
<element type = "[Order] Total Amount" sys_name = "Order_total" sys_type = "Field"/>
<element type = "[Order;Customer] Name" sys_name = "Cust_name" sys_type = "Field"/>
<element type = "[Order;Customer;Address] Address Line 1" sys_name="Cust_address"
sys_type="Field"/>
<element type = "[Order;Customer;Address] City" sys_name = "Cust_city" sys_type = "Field"/>
<element type = "[Order;Customer;Address] Postal Code" sys_name="Cust_pc" sys_type="Field"/>
<element type = "[Order;Customer;Address] Country" sys_name="Cust_country" sys_type="Field"/>
<element type = "[Order;Product] Id" sys_name = "Item1_id" sys_type = "Field"/>
<element type = "[Order;Product] Quantity" sys_name = "Item1_quantity" sys_type = "Field"/>
<element type = "[Order;Product] Price" sys_name = "Item1_price" sys_type = "Field"/>
<element type = "[Order;Product] Id" sys_name = "Item2_id" sys_type = "Field"/>
<element type = "[Order;Product] Quantity" sys_name = "Item2_quantity" sys_type = "Field"/>
<element type = "[Order;Product] Price" sys_name = "Item2_price" sys_type = "Field"/>
</ElementType>
</Schema>
Page 21
Integration Product:
The Integrated Context View
 The
product of the integration is a structurallyneutral hierarchy of concepts called an integrated
context view.
 Define a context view (CV) as follows:
 If
a semantic name Si is in CV, then for any Sj in Si*, Sj is
also in CV.
 For each semantic name Si in CV, there exists a set of zero
or more mappings Mi that associate a schema element Ej
with Si.
 A semantic name Si can only occur once in the CV.
A
context view (CV) is a valid Universal Relation.
 Each
field is assigned a semantic name which uniquely
identifies its semantic connotation.
Page 22
Northwind & Southstorm
Integration Example
Integrated Context View
Integrated View
Term
V (view root)
- [Category]
- Id
- Name
- [Customer]
- Id
- Name
- [Employee]
- Id
- [Name]
- First Name
- Last Name
- [Product]
- Id
- Name
- [Supplier]
- Id
- [Category]
- Id
Data Source Mappings
(not visible to user)
Integrated View
Term
Data Source Mappings
(not visible to user)
N/A
NW.Categories
NW.Categories.CategoryID
NW.Categories.CategoryName
NW.Customers
NW.Customers.CustomerID
NW.Customers.CompanyName
NW.Employees
NW.Employees.EmployeeID
V (view root) (cont.)
- [Order]
-Id
- [Customer]
- Id
- Name
- [Employee]
- Id
- Name
- [Product]
- Id
- Price
- Quantity
- [Shipper]
- Id
- Name
- [Supplier]
- Id
- Name
N/A
NW.Orders, SS.Orders_tb
NW.[Orders,OrderDetails].OrderID, SS.Orders_tb.Order_num
NW.Employees.FirstName
NW.Employees.LastName
NW.Products
NW.Products.PrdouctID
NW.Products.ProductName
NW.Products.SupplierID
NW.Products.CategoryID
NW.Orders.CustomerID
SS.Orders_tb.Cust_name
NW.Orders.EmployeeID
SS.Orders_tb.Emp_name
NW.OrderDetails
NW.OrderDetails.ProductID, SS.Orders_tb.Item[1,2]_id
NW.OrderDetails.UnitPrice, SS.Orders_tb.Item[1,2]_price
NW.OrderDetails.Quantity, SS.Orders_tb.Item[1,2]_qty
NW.Shippers
NW.Shippers.ShipperID
NW.Shippers.ShipperName
NW.Suppliers
NW.Suppliers.SupplierID
NW.Suppliers.SupplierName
Page 23
Architecture Components:
The Query Processor
 The
query processor:
 Allows
the user to formulate queries on the view.
 Translates from semantic names in the context view to
structural queries (SQL) on databases.
 Involves determining correct field and table mappings and
discovery of join conditions and join paths
 Retrieves
query results and formats them for display to
the user.
 Client-side
query processing:
 Perform
joins between databases using common keys.
 Data value formatting and transformation
Page 24
Advanced Query Processing
 Advanced
query processor features include:
 global
keys and joins - a mechanism for specifying
when a field stores a global key such as a social
security number.
 result normalization - a procedure for normalizing
query results returned from each individual database.
(e.g. Southstorm)
 data integration - transforming data representational
conflicts at the global level.
 For example, “M” and “F” may represent “Male” and
“Female” in one database, and another may represent these
concepts using “0” and “1”.
Page 25
Northwind & Southstorm
Query Examples
 Example
1: Retrieve all order ids ([Order] Id) and
customers ([Customer] Name):
 SS:
SELECT Order_num, Cust_name FROM Orders_tb
 NW: SELECT OrderID, CompanyName FROM Orders,
Customers WHERE Orders.CustomerID =
Customers.CustomerID
 Example
2: Retrieve all ordered products
([Order;Product] Id) and their order ids.
 SS:
SELECT Order_num, Item1_id, Item2_id FROM
Orders_tb
 NW: SELECT OrderID, ProductID FROM OrderDetails
 Note: In NW, selects from two different order id mappings.
In SS, result normalization is required.
Page 26
Integration Example:
Discussion
 Important
points:
 System
table and field names are not presented to the
user who queries based on semantic names.
 Database structure is not shown to the user.
 Field and table mappings are automatically
determined based on X-Spec information.
 Join conditions are inserted as needed when available
to join tables.
 Different physical representations for the same
concept are combined.
 Hierarchically related concepts are combined based
on their IS-A relationship in the standard dictionary.
Page 27
Unity Overview
 Unity
is a software package that implements the
integration architecture with a GUI.
 Developed using Microsoft Visual C++ 6 and
Microsoft Foundation Classes (MFC).
 Unity allows the user to:
 Construct
and modify standard dictionaries
 Build X-Specs to describe data sources
 Integrate X-Specs into an integrated view
 Transparently query integrated systems using ODBC
and automatically generate SQL transactions
 Unity
is available for demonstration and
distribution.
Page 28
Architecture Discussion
 The
architecture automatically integrates
relational schemas into a multidatabase.
 Desirable properties:
 Individual
mappings - information sources integrated
one-at-a-time and independently
 Integrated view constructed for query transparency user queries system by semantics instead of structure
 Handles schema conflicts - including semantic,
structural, and naming conflicts
 Automated integration - integrated view constructed
efficiently and automatically
 No wrapper or mediator software is required
 Transparent querying - users issue semantic queries
which are translated to SQL by the query processor Page 31
Contributions
 Architecture
 Has
contributions:
an unique application of a standard dictionary
which is not a knowledge base
 Separates the capture and integration processes
 Produces an integrated, high-level view of all
concepts in the underlying databases
 Allows transparent querying without structure
 Provides algorithms for dynamically extracting
database data (creating relevant views) and for
mediation of global level conflicts
 Arguably simpler method for capturing data semantics
than using description logic
 An implementation, Unity, which demonstrates the
practical benefits of the architecture
Page 32
Conclusions & Future Work
 Automatic
database integration is possible by
using a standard term dictionary and defining
semantic names for schema elements.
 Users are able to transparently query integrated
systems by concept instead of structure.
 We
are constantly refining Unity.
 Develop
an integration component for a web browser
 Test
the system in large industrial projects.
 Allow distributed updates and global updates on
all databases.
Page 33
References
 Publications:
 Unity
- A Database Integration Tool, R. Lawrence and
K. Barker, TRLabs Emerging Technology Bulletin,
January 2000.
 Multidatabase Querying by Context, R. Lawrence and
K. Barker, DataSem2000, pages 127-136, Oct. 2000.
 Integrating Relational Database Schemas using a
Standardized Dictionary, SAC’2001 - ACM Symposium
on Applied Computing, March, 2001.
 Sponsors:
 NSERC,
 Further
TRLabs
Information:
 http://www.cs.umanitoba.ca/~umlawren/
Page 34