Database Reference
In-Depth Information
EETPS. These experiments are classi
ed according to the complexity of the queries into
ve categories: simple, simple-to-medium, medium, medium-to-complex, and complex.
These
ve experiments show comparisons between the response time of retrieving data
from CTTs, VETs, or both CTTs and VETs. The response time of retrieving data from
EET is evaluated by accessing EETPS functions.
6.1 Experimental Setup
EETPS was implemented in Java 1.6.0, Hibernate 4.0, and Spring 3.1.0. The database
is PostgreSQL 8.4 and the application server is Jboss-5.0.0.CR2. Both, the database
and the application server are deployed on the same PC. The operating system is
Windows 7 Home Premium, with Intel Core i5 2.40 GHz CPU, 8 GB of RAM
memory, and 500 GB of hard disk storage.
6.2 Experimental Data Set and Results
EETPS was developed to serve multiple tenants running in a single application
instance, but the aim of the experiments is to evaluate the performance and show the
differences between retrieving data of CTTs, VETs, or both CTTs and VETs together
for a single tenant. As long as in the multi-tenant database, each tenant
'
s data is isolated
in a separate table partition,
these experiments can evaluate the effectiveness of
retrieving each single tenant
s data from EET multi-tenant database. Furthermore, these
experiments are performed using a single server instance, and we do not considered
scale-up or scale-out multi-tenant database issues in this paper. In the
'
ve experiments,
the test is performed on fourteen queries twice, the
rst test to retrieve only 1 row, and
the second test to retrieve 100 rows by using the same queries. In order to produce
accurate comparisons, the same data input is used for CTTs, VETs and CTT-and-VET
to retrieve the same data output. The execution times of these query experiments are
recorded based on six data sets for all the
ve types of experiments. These six data sets
contain, (1) 500 rows, (2) 5,000 rows, (3) 10,000 rows, (4) 50,000 rows, (5) 100,000
rows, and (6) 200,000 rows. In this section, the average execution time is computed by
executing ten tests on each of the six data sets to show accurate results. All of these data
sets were for one tenant. In all the experimental diagrams, the vertical axis shows the
execution time in milliseconds, and the horizontal axis shows the total number of rows
that stored in a tenant
'
s table. In the
ve experiments, the
'
tenant_id
'
equals 1000.
The CTTs that are used in the experiments, including
'
product
'
,
'
sales_fact
'
and
'
'
'
'
of VETs for these tables are 16,
17, and 18 respectively. The data structures used for the queries in the various
experiments are shown in Fig.
15
, and listed below:
(1) Simple Query Experiment
-
Single Table (Exp. 1): In this experiment, the
function of the Single Table Query Algorithm that retrieves data from a CTT is
invoked by executing Query 1 (Q 1) that comprises of the Individual Query (IQ)
1, and retrieve the same data from a VET by executing Query 2 (Q 2) that
comprises of IQ2
sales_details
, and the corresponding
db_table_id
IQ4. These experimental tests show how the Single Table
-
Search WWH ::
Custom Search