A0930105
0 likes545 views
This document discusses techniques to improve the performance of data retrieval from customized data warehouses. It presents a comparison of data retrieval times for queries on relational databases with and without indexing. Indexing techniques like bitmap indexing and data partitioning are shown to improve data retrieval times. Data retrieval times are measured for different sizes of data stored in relational databases and data warehouses, both with and without indexing. The results show that data warehouses with indexing techniques like bitmap indexing provide faster data retrieval times compared to relational databases or data warehouses without indexing, especially as data sizes increase.
![IOSR Journal of Computer Engineering (IOSR-JCE)
e-ISSN: 2278-0661, p- ISSN: 2278-8727Volume 9, Issue 3 (Mar. - Apr. 2013), PP 01-05
www.iosrjournals.org
Performance Improvement Techniques for Customized Data
Warehouse
Md. Al Mamun and Md. Humayun Kabir
Department of Computer Science and Engineering, Jahangirnagar University, Savar, Dhaka-1342, Bangladesh.
Abstract: In this paper, we present performance improvement techniques for data retrieval from customized
data warehouses for efficient querying and Online Analytical Processing (OLAP) in relation to efficient
database and memory management. Different database management techniques, e.g. indexing, partitioning etc.
play vital role in efficient memory management. A comparison of data retrieval time for a particular query from
a relational database as well as data warehouse database with and without indexing is performed. We show that
the application of different database management techniques result faster query execution by reducing data
retrieval time. This improved efficiency may increase the efficiency of OLAP operations, which results better
data warehouse performance.
Keywords - Data Warehouse, Indexing, OLAP, Partitioning, Querying.
I. Introduction
Data retrieval from relational database requires high access time when it stores millions of records,
which can be overcome using indexing [1]. Test data can be generated to populate test databases to test SQL
queries [2]. Different database management techniques e.g. partitioning [3], indexing [1, 4, 5] etc. can be
employed for faster data access. Customized database application software may be developed using these
techniques for faster data processing. In the cases where the number of records in relational database becomes
very high, the query processing time becomes very long [6]. Data warehouses [7-9] which store consolidated
historic data can be constructed from relational databases. Data warehouse (DW) database store very small
number of records for a large number of records in relational database [6]. This reduction of the number of
records in data warehouse results in smaller query retrieval time [6]. Moreover, this retrieval time can be further
reduced using different indexing techniques.
In this paper, we have studied different techniques to improve performance of data retrieval from
relational database as well as DW [8, 9] database for different sizes of data. We have used bitmap indexing
(BMI) [10] and data partitioning techniques to improve performance of data retrieval from data warehouse
database [6]. We have also shown that data retrieval time for data warehouse is very much lower compared to
relational database for similar queries using bitmap indexing [10]. This suggests that data warehouse with
bitmap indexing is more suitable for an enterprise for intelligent and faster decision making [3, 6]. The retrieval
time rises with the increase in data size.
The paper is organized as follows. Section II presents system architecture, section III describes the
data retrieval time measurement technique, section IV presents comparison of data retrieval time for RDBMS
with and without indexing, section V presents about data retrieval time for partitioning, section VI presents
comparison of data retrieval time for different data sources with variable data sizes, section VII concludes.
II. System Architecture
This section presents the architecture for constructing customized data warehouses from RDBMS
tables using data definition language (DDL) of SQL. Data warehouse is populated with data from external
sources e.g. relational database system, in the consolidated form using aggregation operation. We have created
dimension and fact tables from RDBMS table schemas. Queries are executed on both RDBMS tables and DW
dimension and fact tables with and without indexing using Java programs. The data retrieval time are measured
and compared. Fig 1 represents the system architecture.
III. Data Retrieval Time
We have designed an algorithm for determining data retrieval time for relational database system and DW
database system [6]. We have used the function DBTime () to determine data retrieval time: tr = (tend – tstart ) in
milliseconds (ms) of executing a particular query.
www.iosrjournals.org 1 | Page](https://image.slidesharecdn.com/a0930105-130401063313-phpapp01/85/A0930105-1-320.jpg)
![Performance Improvement Techniques for Customized Data Warehouse
Data Sources RDB as input
RDBMS Java Program
utput SQL Query
Query
DDL on RDB Schema
Execution
Dimension and Fact Tables
Data
Warehouse
Populate
Data Cube Populated DW as Input
Data Warehouse
RDBMS
RDB Output
Visual Query
Text/Graphical Output Function Execution
DW Output
Figure 1: System Architecture
Retrieval time varies with primary memory and processing speed of computer in which we execute our software
system. The experiment is done on a laptop with Core i3 processor of 2.53 GHz, RAM 2GB, HD 500GB under
windows OS.
IV. Comparison Of Data Retrieval Time For Relational Databases With And Without
Indexing
We have applied indexing on students records stored in database relations using Oracle RDBMS. The
developed performance improvement system generates very large number of student random data records by
varying their CGPAs to populate the RDBMS tables. The developed prototype determines the data retrieval time
using data tables created of different data sizes using RDBMS for a particular query without indexing at first as
shown in Table 1 [6].
Table 1 Data retrieval time without indexing and with indexing on tables of different data sizes.
Number of The database relations are then indexed and the
Records Retrieval Time (ms) same query is executed on the indexed relations. The
Data size in Without With indexing query execution times are recorded as shown in Table
number of indexing
1. Fig 2 plots query execution time without and with
records
indexing on different relations of variable data sizes.
100000 3324 2819
200000 10671 5457 We notice that the data retrieval time for non-indexed
300000 47727 35340 relations is more than that of indexed relations for a
particular data size.
Figure 2: Comparison of data retrieval time without indexing and with indexing for RDBMS.
V. Data Retrieval Time Using Data Partitioning
Data partitioning can speed up the performance of data processing in data retrieval. In case of small
physical memory, the large volume of data can be partitioned into smaller segments to load into primary
memory. This can help to execute the application program to access data larger than the main memory size
successfully. But if the number of partitions is big enough, then data access
www.iosrjournals.org 2 | Page](https://image.slidesharecdn.com/a0930105-130401063313-phpapp01/85/A0930105-2-320.jpg)
![Performance Improvement Techniques for Customized Data Warehouse
may take longer time due to switching between partitions as overhead [3]. Increasing physical memory size,
partitioning can be avoided or number of partitions can be reduced resulting faster data access. Partitioning with
indexing may cause even more reduction in data retrieval time.
VI. Comparison Of Data Retrieval Time For Different Data Sizes
Consider the data retrieval time shown in Table 2 required to select 209 students of a particular session
2002-2003, who obtained CGPA 4.0 out of 100000 student records by executing a query. We have defined and
executed queries to retrieve records of Table 3 from RDBMS, DW without and with bitmap indexing.
Table 2 Data retrieval time (in ms) for RDBMS, Data Warehouse without indexing, and Data Warehouse with
BMI
RDBMS DW (non-
(indexed) indexed) access DW with BMI
access time time access time
92 41 28
Retrieval Time for RDBMS and DW
100
Data Retrieval Time
80
60
Series1
40
20
0
RDBMS Data Warehouse Data Warehouse
with BMI
Data Sources
Figure 3: Retrieval time for indexed RDBMS, Data Warehouse without indexing, and Data Warehouse with BMI for
selecting 209 students of a session.
Consider the queries to retrieve session, final exam year, CGPA, and the number of the students who
obtained CGPA 4 as shown in Table 3. The queries retrieve and count student’s records which are stored in
RDBMS and DW. Table 4 represents data access time in ms for indexed RDBMS, non-indexed DW and DW
with bitmap indexing.
TABLE 3 Query Output for students of all sessions with CGPA 4.0
4th Year No. of
Session Final Exam CGPA Students
2001-2002 2005 4 503
2002-2003 2006 4 209
2003-2004 2007 4 236
2004-2005 2008 4 265
2005-2006 2009 4 241
2006-2007 2010 4 239
2007-2008 2011 4 261
TABLE 4 Data retrieval time for students of all sessions with CGPA 4.0
Indexed Data Data Warehouse
RDBMS Warehouse with BMI
555 223 207
Fig 4 plots the data retrieval time shown in Table 4 for different data sources to retrieve the query output
shown in Table 3.
www.iosrjournals.org 3 | Page](https://image.slidesharecdn.com/a0930105-130401063313-phpapp01/85/A0930105-3-320.jpg)
![Performance Improvement Techniques for Customized Data Warehouse
Figure 5: Comparison of data retrieval time for different data Sizes shown in Table 7.
TABLE 8 Retrieval Time for Data Warehouses of different Data Sizes.
DW Retrieval Time vs Data Size
Data size of DW Retrieval DW Retrieval
DW Time Time with BMI
60000
1954 223 207
50000
5879 538 502
Retrieval Time
40000
9880 769 435
30000
19986 8569 7520
20000
10000
40060 17927 17222
0
79803 53057 38213
1954 5879 9880 19986 40060 79803
Data Size
DW Retrieval Time DW Retrieval Time w ith BMI
Figure 6: Comparison of data retrieval time for different data Sizes of Data Warehouse with and without BMI shown in
Table 8.
Fig. 5 and Fig. 6 explain that data retrieval time in both RDBMS and DW cases increase with the increase in data size
significantly.
VII. Conclusion
We have observed that data retrieval from tables stored in RDBMS database is almost exponentially
rising. But the increases in data retrieval time for data warehouse with bitmap indexing or without bitmap
indexing have small increase with the increase of data size. We have shown that data retrieval time for data
warehouse is very much lower compared to relational database for similar queries. This suggests that data
warehouse is more suitable for an enterprise for intelligent and faster data access in decision making. This
suggests that OLAP system can be developed using DW database and is more suitable than using relational
database system for intelligent and efficient decision making with reporting or data analysis.
Acknowledgements
We greatly acknowledge the valuable comments of the faculty members who were present at the thesis
examination board.
References
[1] M. Barrena, C. Pachon and E. Jurado, JISBD2007-04: Neighbors search in holey multidimensional spaces, IEEE Latin America
Transactions, Vol. 6, No. 4, Aug. 2008, pages 332-338.
[2] M. J. Suarez-Cabal, C. de la Riva and J. Tuya, JISBD04-Populating Test Databases for Testing SQL Queries, IEEE Latin America
Transactions, Vol. 8, No. 2, April 2010, pages 164-171.
[3] Mafruz Zaman Ashrafi, David Taniar, Kate Smith, ODAM: An Optimized Distributed Association Rule Mining Algorithm, Monash
University, IEEE Distributed Systems Online, IEEE Computer Society, Vol. 5, No. 3; March 2004.
[4] Bernd Reiner, Karl Hahn. Optimized Management of Large-Scale Data Sets Stored on Tertiary Storage System, IEEE Distributed
Systems Online, IEEE Computer Society, Vol. 5, No. 5; May 2004.
[5] S. Repp, A. Gross and C. Meinel, Browsing within Lecture Videos based on the chain index of speech transcription, IEEE
transactions on learning technologies, Vol. 1, Issue 3, 2008, pages 145-156.
[6] M. Al Mamun. Data Warehouse Performance Analysis for Online Analytical Processing, A Thesis Draft submitted for predefense of
MS, Department of Computer Science and Engineering, Jahangirnagar University, Savar, Dhaka-1342.
[7] V. Nebot and R. Berlanga, JISBD02-Populating Data Warehouses with Semantic Data, IEEE Latin America Transactions, Vol. 8,
No. 2, April 2010, pages 150-157.
[8] J.-N. Mazon and J. Trujillo, JISBD2007-02: Model-driven reverse engineering for data warehouse design, IEEE Latin America
Transactions, Vol. 6, No. 4, Aug. 2008, pages 317-323.
[9] E. Soler, J. Trujillo, E. Fernandez-Medina, and M. Piattini, JISBD2007-07: An extension of the relational metamodel of CWM to
represent secure data warehouse at the logical level, IEEE Latin America Transactions, Vol. 6, No. 4, Aug. 2008, pages 355-362.
[10] Morteza Zaker, Somnuk Phon-Amnuaisuk and Su-Cheng Haw, An Adequate Design for Large Datawarehouse systems: Bitmap
indexes versus B-tree index, International Journal of Computers and Communications, Vol. 2, Issue 2, 2008.
www.iosrjournals.org 5 | Page](https://image.slidesharecdn.com/a0930105-130401063313-phpapp01/85/A0930105-5-320.jpg)
A0930105
- 1. IOSR Journal of Computer Engineering (IOSR-JCE) e-ISSN: 2278-0661, p- ISSN: 2278-8727Volume 9, Issue 3 (Mar. - Apr. 2013), PP 01-05 www.iosrjournals.org Performance Improvement Techniques for Customized Data Warehouse Md. Al Mamun and Md. Humayun Kabir Department of Computer Science and Engineering, Jahangirnagar University, Savar, Dhaka-1342, Bangladesh. Abstract: In this paper, we present performance improvement techniques for data retrieval from customized data warehouses for efficient querying and Online Analytical Processing (OLAP) in relation to efficient database and memory management. Different database management techniques, e.g. indexing, partitioning etc. play vital role in efficient memory management. A comparison of data retrieval time for a particular query from a relational database as well as data warehouse database with and without indexing is performed. We show that the application of different database management techniques result faster query execution by reducing data retrieval time. This improved efficiency may increase the efficiency of OLAP operations, which results better data warehouse performance. Keywords - Data Warehouse, Indexing, OLAP, Partitioning, Querying. I. Introduction Data retrieval from relational database requires high access time when it stores millions of records, which can be overcome using indexing [1]. Test data can be generated to populate test databases to test SQL queries [2]. Different database management techniques e.g. partitioning [3], indexing [1, 4, 5] etc. can be employed for faster data access. Customized database application software may be developed using these techniques for faster data processing. In the cases where the number of records in relational database becomes very high, the query processing time becomes very long [6]. Data warehouses [7-9] which store consolidated historic data can be constructed from relational databases. Data warehouse (DW) database store very small number of records for a large number of records in relational database [6]. This reduction of the number of records in data warehouse results in smaller query retrieval time [6]. Moreover, this retrieval time can be further reduced using different indexing techniques. In this paper, we have studied different techniques to improve performance of data retrieval from relational database as well as DW [8, 9] database for different sizes of data. We have used bitmap indexing (BMI) [10] and data partitioning techniques to improve performance of data retrieval from data warehouse database [6]. We have also shown that data retrieval time for data warehouse is very much lower compared to relational database for similar queries using bitmap indexing [10]. This suggests that data warehouse with bitmap indexing is more suitable for an enterprise for intelligent and faster decision making [3, 6]. The retrieval time rises with the increase in data size. The paper is organized as follows. Section II presents system architecture, section III describes the data retrieval time measurement technique, section IV presents comparison of data retrieval time for RDBMS with and without indexing, section V presents about data retrieval time for partitioning, section VI presents comparison of data retrieval time for different data sources with variable data sizes, section VII concludes. II. System Architecture This section presents the architecture for constructing customized data warehouses from RDBMS tables using data definition language (DDL) of SQL. Data warehouse is populated with data from external sources e.g. relational database system, in the consolidated form using aggregation operation. We have created dimension and fact tables from RDBMS table schemas. Queries are executed on both RDBMS tables and DW dimension and fact tables with and without indexing using Java programs. The data retrieval time are measured and compared. Fig 1 represents the system architecture. III. Data Retrieval Time We have designed an algorithm for determining data retrieval time for relational database system and DW database system [6]. We have used the function DBTime () to determine data retrieval time: tr = (tend – tstart ) in milliseconds (ms) of executing a particular query. www.iosrjournals.org 1 | Page
- 2. Performance Improvement Techniques for Customized Data Warehouse Data Sources RDB as input RDBMS Java Program utput SQL Query Query DDL on RDB Schema Execution Dimension and Fact Tables Data Warehouse Populate Data Cube Populated DW as Input Data Warehouse RDBMS RDB Output Visual Query Text/Graphical Output Function Execution DW Output Figure 1: System Architecture Retrieval time varies with primary memory and processing speed of computer in which we execute our software system. The experiment is done on a laptop with Core i3 processor of 2.53 GHz, RAM 2GB, HD 500GB under windows OS. IV. Comparison Of Data Retrieval Time For Relational Databases With And Without Indexing We have applied indexing on students records stored in database relations using Oracle RDBMS. The developed performance improvement system generates very large number of student random data records by varying their CGPAs to populate the RDBMS tables. The developed prototype determines the data retrieval time using data tables created of different data sizes using RDBMS for a particular query without indexing at first as shown in Table 1 [6]. Table 1 Data retrieval time without indexing and with indexing on tables of different data sizes. Number of The database relations are then indexed and the Records Retrieval Time (ms) same query is executed on the indexed relations. The Data size in Without With indexing query execution times are recorded as shown in Table number of indexing 1. Fig 2 plots query execution time without and with records indexing on different relations of variable data sizes. 100000 3324 2819 200000 10671 5457 We notice that the data retrieval time for non-indexed 300000 47727 35340 relations is more than that of indexed relations for a particular data size. Figure 2: Comparison of data retrieval time without indexing and with indexing for RDBMS. V. Data Retrieval Time Using Data Partitioning Data partitioning can speed up the performance of data processing in data retrieval. In case of small physical memory, the large volume of data can be partitioned into smaller segments to load into primary memory. This can help to execute the application program to access data larger than the main memory size successfully. But if the number of partitions is big enough, then data access www.iosrjournals.org 2 | Page
- 3. Performance Improvement Techniques for Customized Data Warehouse may take longer time due to switching between partitions as overhead [3]. Increasing physical memory size, partitioning can be avoided or number of partitions can be reduced resulting faster data access. Partitioning with indexing may cause even more reduction in data retrieval time. VI. Comparison Of Data Retrieval Time For Different Data Sizes Consider the data retrieval time shown in Table 2 required to select 209 students of a particular session 2002-2003, who obtained CGPA 4.0 out of 100000 student records by executing a query. We have defined and executed queries to retrieve records of Table 3 from RDBMS, DW without and with bitmap indexing. Table 2 Data retrieval time (in ms) for RDBMS, Data Warehouse without indexing, and Data Warehouse with BMI RDBMS DW (non- (indexed) indexed) access DW with BMI access time time access time 92 41 28 Retrieval Time for RDBMS and DW 100 Data Retrieval Time 80 60 Series1 40 20 0 RDBMS Data Warehouse Data Warehouse with BMI Data Sources Figure 3: Retrieval time for indexed RDBMS, Data Warehouse without indexing, and Data Warehouse with BMI for selecting 209 students of a session. Consider the queries to retrieve session, final exam year, CGPA, and the number of the students who obtained CGPA 4 as shown in Table 3. The queries retrieve and count student’s records which are stored in RDBMS and DW. Table 4 represents data access time in ms for indexed RDBMS, non-indexed DW and DW with bitmap indexing. TABLE 3 Query Output for students of all sessions with CGPA 4.0 4th Year No. of Session Final Exam CGPA Students 2001-2002 2005 4 503 2002-2003 2006 4 209 2003-2004 2007 4 236 2004-2005 2008 4 265 2005-2006 2009 4 241 2006-2007 2010 4 239 2007-2008 2011 4 261 TABLE 4 Data retrieval time for students of all sessions with CGPA 4.0 Indexed Data Data Warehouse RDBMS Warehouse with BMI 555 223 207 Fig 4 plots the data retrieval time shown in Table 4 for different data sources to retrieve the query output shown in Table 3. www.iosrjournals.org 3 | Page
- 4. Performance Improvement Techniques for Customized Data Warehouse Comparison of Data Retrieval Time 600 500 Retrieval Time 400 300 Series1 200 100 0 RDBMS DW DW with BMI Data Sources Figure 4: Comparison of data retrieval time for different data sources without and with indexing Queries operated on databases of RDBMS require the highest time as it stores a large number of raw data records. Table 5 represents the data retrieval time of executing a query on data tables of various sizes containing records of up to 4 millions using RDBMS and the corresponding records in DW. We have measured the execution time of a query for accessing data from an indexed RDBMS database, non-indexed data warehouse database and a DW with bitmap indexing. It is observed that data retrieval from data warehouse with bitmap indexing requires less time compared to that of data warehouse without indexing. TABLE 5 Retrieval time for CGPA 4.0 students of all sessions from indexed data tables of different sizes stored in RDBMS, DW without and with bitmap indexing RDBMS Data Retrieval Time (ms) No. of Indexed DW with Records RDBMS DW BMI 100000 555 223 207 300000 8964 538 502 500000 13000 769 435 1000000 16259 8569 7520 2000000 42546 17927 17222 4000000 87028 53057 38213 We create two tables Table 6 and Table 7 based on Table 5. Finally, we create another two tables Table 7 and Table 8 for clarification of the data retrieval time for different data sizes of RDBMS database and DW database separately. TABLE 6 Corresponding Data Sizes of DW database for RDBMS database Data Size of RDBMS Data size in DW 100000 1954 300000 5879 500000 9880 1000000 19986 2000000 40060 4000000 79803 TABLE 7 Retrieval Time for different Data Sizes of RDBMS RDBMS RDBMS Retrieval Data Size Time 100000 555 300000 8964 500000 13000 1000000 16259 2000000 42546 4000000 87028 www.iosrjournals.org 4 | Page
- 5. Performance Improvement Techniques for Customized Data Warehouse Figure 5: Comparison of data retrieval time for different data Sizes shown in Table 7. TABLE 8 Retrieval Time for Data Warehouses of different Data Sizes. DW Retrieval Time vs Data Size Data size of DW Retrieval DW Retrieval DW Time Time with BMI 60000 1954 223 207 50000 5879 538 502 Retrieval Time 40000 9880 769 435 30000 19986 8569 7520 20000 10000 40060 17927 17222 0 79803 53057 38213 1954 5879 9880 19986 40060 79803 Data Size DW Retrieval Time DW Retrieval Time w ith BMI Figure 6: Comparison of data retrieval time for different data Sizes of Data Warehouse with and without BMI shown in Table 8. Fig. 5 and Fig. 6 explain that data retrieval time in both RDBMS and DW cases increase with the increase in data size significantly. VII. Conclusion We have observed that data retrieval from tables stored in RDBMS database is almost exponentially rising. But the increases in data retrieval time for data warehouse with bitmap indexing or without bitmap indexing have small increase with the increase of data size. We have shown that data retrieval time for data warehouse is very much lower compared to relational database for similar queries. This suggests that data warehouse is more suitable for an enterprise for intelligent and faster data access in decision making. This suggests that OLAP system can be developed using DW database and is more suitable than using relational database system for intelligent and efficient decision making with reporting or data analysis. Acknowledgements We greatly acknowledge the valuable comments of the faculty members who were present at the thesis examination board. References [1] M. Barrena, C. Pachon and E. Jurado, JISBD2007-04: Neighbors search in holey multidimensional spaces, IEEE Latin America Transactions, Vol. 6, No. 4, Aug. 2008, pages 332-338. [2] M. J. Suarez-Cabal, C. de la Riva and J. Tuya, JISBD04-Populating Test Databases for Testing SQL Queries, IEEE Latin America Transactions, Vol. 8, No. 2, April 2010, pages 164-171. [3] Mafruz Zaman Ashrafi, David Taniar, Kate Smith, ODAM: An Optimized Distributed Association Rule Mining Algorithm, Monash University, IEEE Distributed Systems Online, IEEE Computer Society, Vol. 5, No. 3; March 2004. [4] Bernd Reiner, Karl Hahn. Optimized Management of Large-Scale Data Sets Stored on Tertiary Storage System, IEEE Distributed Systems Online, IEEE Computer Society, Vol. 5, No. 5; May 2004. [5] S. Repp, A. Gross and C. Meinel, Browsing within Lecture Videos based on the chain index of speech transcription, IEEE transactions on learning technologies, Vol. 1, Issue 3, 2008, pages 145-156. [6] M. Al Mamun. Data Warehouse Performance Analysis for Online Analytical Processing, A Thesis Draft submitted for predefense of MS, Department of Computer Science and Engineering, Jahangirnagar University, Savar, Dhaka-1342. [7] V. Nebot and R. Berlanga, JISBD02-Populating Data Warehouses with Semantic Data, IEEE Latin America Transactions, Vol. 8, No. 2, April 2010, pages 150-157. [8] J.-N. Mazon and J. Trujillo, JISBD2007-02: Model-driven reverse engineering for data warehouse design, IEEE Latin America Transactions, Vol. 6, No. 4, Aug. 2008, pages 317-323. [9] E. Soler, J. Trujillo, E. Fernandez-Medina, and M. Piattini, JISBD2007-07: An extension of the relational metamodel of CWM to represent secure data warehouse at the logical level, IEEE Latin America Transactions, Vol. 6, No. 4, Aug. 2008, pages 355-362. [10] Morteza Zaker, Somnuk Phon-Amnuaisuk and Su-Cheng Haw, An Adequate Design for Large Datawarehouse systems: Bitmap indexes versus B-tree index, International Journal of Computers and Communications, Vol. 2, Issue 2, 2008. www.iosrjournals.org 5 | Page