|
|
Thesis: "A Clustering
Approach to Configuring Buffer Pools in a Database Management System"
|
|
Abstract (full text
not available)
|
|
Database Management Systems
(DBMSs) use a main memory area as a buffer to reduce the number of disk accesses
performed by a transaction. This memory is a key resource in a DBMS.
Current DBMSs such as DB2 Universal Database, divide the buffer area
into a number of independent buffer pools. Each database object (table
or index) is assigned to a specific buffer pool. The tasks of configuring
the buffer pools, which defines the mapping of database objects to buffer
pools and setting a size for each of the buffer pools, is crucial for achieving
optimal performance.
Mapping database objects to buffer pools, which we refer to as "the buffer
pool configuration problem", is the focus of this thesis research. Mapping
database objects to buffer pools can be viewed as a partitioning problem,
that is, we partition the database objects into groups where each group goes
to one buffer pool. The partitioning of objects is based on how the
objects are used and on the inherent properties of objects.
This thesis presents an approach to the configuration problem based on analyzing
access patterns to database objects in a given database workload. The
approach is discussed, implemented and verified against the TPC-C benchmark
database. The performance metrics used to evaluate the algorithms are the
weighted response time, throughput, and percentage of physical reads for
the given workload.
|
|