Databases Reference
In-Depth Information
fragmentation schemas into a chromosomes. (2) Define the fitness function. (3)
Select optimal fragmentation schema by Genetic Algorithms : To realize the
GA
selection, we use a JAVA API called
JGAP
2
(Java Genetic Algorithms Package)
that implements genetic algorithms. JGAP needs two inputs: the chromosome
encoding and the fitness function, it gives in output the optimal (near-optimal)
fragmentation schema. The JGAP API is based on a
GA
process :
GA
generate
an initial population of chromosomes, then performs genetic operations (selec-
tion, mutation, crossover) in order to generate new populations. Each chromo-
some is evaluated by fitness function in order to estimate the benefit given by
the fragmentation schema to the workload performance. The process of
HDP
selection by GA is given as follow:
HDP
selection by
GA
Input:
Q
: workload of
t
queries
AS
:
n
fragmentation attributes
AS
=
{A
1
, ··· ,A
n
}
Dom
: attributes domains
Dom
=
{Dom
1
, ··· ,Dom
n
}
DW
: Data of the cost model (table size, system page etc.)
B
: maintenance bound given by Administrator (maximum number of fact fragments)
Output:
fragmentation schema
SF
.
Notations:
Chrom Encoding
: Encode a fragmentation schema into a chromosome
F itnessHDP
: fitness function for the
GA
JGAP
: JAVA API that implements the
GA
Begin
ChromosomeHDP
=
Chrom Encoding
(
AS, Dom
);
F itnessHDP
=
Genetic F itnessFonction
(
Q,AS,B,DW
);
SF
=
JGAP
(
ChromosomeHDP, F itnessHDP
);
End
4 Incremental Selection of Data Warehouse
Fragmentation Schemas Using Genetic Algorithms
In the context of physical design of data warehouse, many studies are focused on
HDP
selection problem in order to define fragmentation schema that improves a
workload performance. But the majority of these works define a static approach
that can't deal with changes occurring on
DW
, specially the execution of news
queries that not exist in the current workload. To achieve incremental selection
of a fragmentation schema, we must adjust the current fragmentation schema
of
DW
by taking into account the execution of a new query
Q
i
. Running
Q
i
Search WWH ::
Custom Search