New Concepts for Parallel Object-Relational Query Processing

"During the last few years, parallel object-relational database management systems have emerged as the leading data management technology on the market. These systems are extensible by user-defined data types and user-defined functionality for the data.
This work focuses on the efficient parallel execution of user-defined functionality. The main contributions describe techniques to support data parallelism for user-defined scalar and aggregate functions and intra-function parallelism for the execution of a scalar function on a large object, and a new technology to provide extensibility with regard to new set-oriented database operations that can efficiently implement user-defined functionality in parallel object-relational database management systems. "

CHAPTER 1 Introduction
1.1 ORDBMS：The Next Great Wave
1.2 Extensible DBMS
1.3 Overview
CHAPTER 2 Background on User-Defined Routines
2.1 User-Defined Routines
2.2 Definition，Implementation，and Execution ofNew UDR
2.2.1 User.Defined Scalar Functions
2.2.2 User-Defined Aggregate Functions
2.2.3 User.Defined Table Functions
2.2.4 User-Defined Functions and Large Objects
2.3 Comparison with Stored Procedures
2.4 Optimization of Queries with UDF
CHAPTER 3 Parallel Processing of User-Defined Functions
3.1 Introduction
3.2 Limits of Current ORDBMS
3.3 Parallel Processing Of UDF
3.3.1 Two Step Parallel Aggregation Of UDAF
3.3.2 Partitioning Classes and Partitionable Functions
3.3.3 Parallel Sorting as a Preprocessing Step for UDAF
3.3.4 Extended Syntax for Function Registration
3.4 Example Applications
3.4.1 The UDAF Most Frequent
3.4.2 The UDSF Running-Average
3.4.3 The UDAF Median
3.4.4 Further Applications
3.5 Plausibility Considerations Regarding Performance
3.6 Related Work
3.7 Summary
CHAPTER 4 Intra-function Parallelism
4.1 Introduction
4.2 Compose/Decompose Operators for Intra-function Parallelism
4.2.1 Compose／Decompose Operators
4.2.2 Extensibility of Compose Operators by Combine Functions
4.2.3 Application of Intra.function Parallelism
4.2.4 Intra-function Parallelism for Function Pipelines
4.3 Experimental Performance Study
4.3.1 Experimental Scenario and Implementation
4.3.2 Performance Results
4.4 Related WOrk
4.5 Summary
CHAPTER 5 The Multi-operator Method
5.1 Introduction
5.2 Performance Problems with Complex UDF in Current ORDBMS
5.2.1 The PBSM Algorithm as a Sophisticated UDP Implementation
5.3 The Multi.operator Method as a New Technique to Implement Complex
5.3.1 The Multi—operator Method and Its Benefits
5.3.2 A Multi.operator Implementation of the PB SM Algorithm
5.4 Supporting the Multi.operator Method
5.4.1 Executing Query Execution Plans
5.4.2 Example for a Textual Specification of Query Execution Plans
5.4.3 Parallel Evaluation
5.5 Performance Evaluation
5.5.1 Experimental Scenario
5.5.2 Performance Results
5.6 Related Work
5.7 Summary
CHAPTER 6 User-Deftned Table Operators
6.1 Introduction.
6.2 User-Defined Table Operators
6.2.1 A Generalization Relationship for Row Types
6.2.2 Defining and Implementing UDTO
6.2.3 The Difrerent Usages of the UDTO Concept
6.2.4 Parallel Processing of Procedural UDTO
6.2.5 Extension to Multiple OUtput Tables
6.3 Example Applications for UDTO
6.3.1 Computing a Spatial Join
6.3.2 Different UDT0 for the Same Predicate
6.3.3 Computing the Median：An Aggregation Operator
6.3.4 A UDTO for a Complex Aggregation
6.3.5 Association Rule Mining
6.4 Related Work
6.5 Summary and Conclusions
CHAPTER7 Implementation ofUDTO
7.1 Introduction
7.2 The MIDAS Prototype
7.2.1 Architectural Overview
7.2.2 Query Compilation and Execution
7.2.3 The MIDAS System Tables
7.2.4 UDSF in MIDAS
7.3 Implementation of SOL Macros
7.3.1 DDL Statements
7.3.2 SOL Macro Expansion in DML Statements
7.3.3 Expanding SQL Macros in Preprocessors and Middleware
……
CHAPTER 8 Summary，Conclusions，and Future Work
References
AppendixA