|
|
Thesis: "Automatically
Scheduling Database Utilities"
|
|
Abstract (full text
not available)
|
|
Current database management
systems (DBMSs) require careful database configuration and maintenance for
good performance. Database system administration is a complex task,
often requiring expert knowledge of database design and application behaviour.
Expert database administrators are required to configure and maintain
an enterprise-class database, which increases the cost of DBMS ownership.
Automating the administration tasks is crucial to making DBMSs more
affordable.
Data maintenance is one of the important administrative tasks in DBMSs.
Database administrators use data maintenance utilities to ensure the
data in tables are stored as efficiently as possible. Without data
maintenance, the performance of applications will decrease but executing
maintenance utilities interferes with running applications. Scheduling
and executing the utilities properly is therefore crucial to preventing applications
from suffering major performance degradation.
The aim of this thesis is to find a usable algorithm to automatically schedule
database utilities so that the performance degradation of applications is
minimal. The scheduling problem is formulated as a combinatorial optimization
problem. Four local (or heuristic) search algorithms, greedy search,
simulated annealing, tabu search, and genetic algorithm are implement and
evaluated for solving the database utilities scheduling problem. Experimental
results show that the tabu search algorithm is the best algorithm in terms
of the quality of the schedule, the level of consistency, and scheduling speed.
The effectiveness of the tabu search algorithm is future demonstrated
by also solving two large "real-world" database utilities scheduling problems.
|
|