You are on page 1of 6

BAHIR DAR UNIVERSITY

INSTITUTE OF TECHNOLOGY
COMPUTING FACULTY
Department of Information Technology
Post Graduate Program
Distributed Database System

A Review on: “Optimal Ant and Join Cardinality for Distributed Query
Optimization Using Ant Colony Optimization Algorithms”
Authors: “Preeti Tiwari and Swati V. Chande”

Reviewed by: 1. Kalkidan Alayu BDU1401430PR


2. Rediet Yimenu BDU1402022PR

Submitted To: Dr. Abdulkerim M


Submitted Date: July, 2022
Table of Contents
A Review on: “Optimal Ant and Join Cardinality for Distributed Query Optimization
Using Ant Colony Optimization Algorithms” ............................................................................ 0

1. Introduction ........................................................................................................................... 2

2. Objective ................................................................................................................................. 2

3. Methodology ........................................................................................................................... 2

3.1. Techniques ........................................................................................................................................ 2


3.2. Dataset.......................................................................................................................................... 3
4. Results and analysis ............................................................................................................... 3

5. Critiques ................................................................................................................................. 4

5.2. Strong side ................................................................................................................................... 4


5.3. Weakness ..................................................................................................................................... 4
6. Conclusion .............................................................................................................................. 4

7. References ............................................................................................................................... 5

Page | 1
1. Introduction
The Distributed Database Management System (DDBMS) offers higher data availability,
reliability, and security to data and enhances the performance and efficiency of the system by
implementing Transparency and Local Autonomy.(Connolly & Begg, 2007)

With the aid of Local and Global Database Manager, a query is stored across various sites in a
distributed system. The query is parsed, validated, and decomposed into algebraic form based on
global conceptual schema as it enters the DBMS environment. In this case, the localization
manager reorganizes the local schema and creates fragmented queries.

Many equivalent query execution plans are generated with some better than others in terms of
execution cost and execution time by permuting the order of operations. These equivalent query
execution plans are received by Global Distributed Query Optimizer (GDQO) that evaluates all
permutated query plans based on data transmission statistics between sites (Liu & Yu, 1993).

The major objective of GDQO is to find the “best” query execution strategy among the candidate
solutions to ensure quick, effective, accurate, and reliable results with minimum utilization of the
system resources.

Ant Colony Optimization Algorithm (ACO) has received a lot of interest as a GDQO to handle
query optimization problems in RDBMS and DDBMS. This is due to the algorithm's constructive
approach of evaluating each alternative dynamically at the time of execution and creating better
plans rather than first creating a plan and then evaluating and improving.

2. Objective
The major objective of this paper is to determine the minimum number of ants needed to optimize
a distributed query with varied number of joins.

3. Methodology
3.1. Techniques
To implement ACO as an optimizer in a DDBMS environment, the query is represented as a
connected construction graph G = (N, E) where N denotes node representing the relations involved

Page | 2
in query and E denotes the edges representing the join between these relations because ACO works
best when problem is represented in the form of graphs. The length of the edges connecting the
vertices is equivalent to the cardinality of the relations and each vertex is associated with the
pheromone values and heuristics values.
With the least amount of ants possible, the simulation system is made to optimize enormously
massive join queries in distributed databases. The method provides a search space that comprises
of many ways to carry out the join operation depending on how many joins are included in the
fragmented query. The ants seek for, build, and extract the best Join Order that involves the fewest
number of tuples in the shortest amount of time. The overall number of ants executing this search
process is important. The method examines the Optimal Cardinality of the Join Operation as well
as the time required for evaluation and production of the optimized Join Order.
3.2. Dataset
The method accepts as input a number of joins ranging from 25 to 100 with a 5 interval. ACO is
performed using a variable number of ants ranging from 1 to 30 with each join number (with an
interval of 1). The algorithm runs for varying numbers of ants, from one ant solving the query
optimization problem to 30 ants solving the same query optimization problem, for each query
executing join operation involving x number of joins.

4. Results and analysis


The Number of Ants (NOA) taken for 25 and 30 joins is 1–25. For Number of Joins (NOJ) 35, 40,
45, 50, and 55, NOA is set to 1–30. Observations state that when the number of ants is extremely
less, i.e., one to three, the algorithm tends to fall into local optimum exhibiting premature
convergence.

Table 1 Minimum ants analysis optimal solutions

Page | 3
When the number of ants is extremely low, minimum time is taken to generate optimized result
sets. The analysis also indicates that as the number of ant increases, the Cardinality decreases.

Improvement Percentage (IP) estimates the improvement in Cardinality of the join operation when
number of ants increases from one to estimation of optimized results.

As the result table shows the number of Joins changes in the distributed query, the number of ants
needed to generate optimal result also changes.

Hence, to estimate minimum number of ants needed to generate optimized results, Ant Ratio is
calculated.

Observations on evaluation of Ant Ratio states that optimal results are obtained when the ratio falls
in the range 0.2–0.3 or the optimal number of ants should be 1/5th–3/10th of the problem size.

5. Critiques
5.2. Strong side
▪ The greedy heuristic aids in the early stages of the Search process in finding acceptable
solutions.
▪ Distributed computing prevents early convergence
5.3. Weakness
▪ Slower convergence compared to other Heuristics
▪ There is no central processor to direct the ant colony to effective solutions.
▪ The Authors didn’t mention about future research areas.

6. Conclusion
This study focuses on estimating the smallest ant population required to optimize distributed
queries using the Ant Colony Optimization Algorithm. The study of the query results considers
both Join Cardinality and Response Time. According to experimental findings, as the ant

Page | 4
population grows, the execution time also grows as a result of the ants' initial lack of systematic
pheromone-induced exploration. However, the Join Cardinality gets better as the ant population
grows.

7. References
Connolly, T., & Begg, C. (2007). Database systems-a practical approach to design,
implementation and management 3rd ed. Pearson Education.
Liu, C., & Yu, C. (1993). Performance Issues in Distributed Query Processing. IEEE Transactions
on Parallel and Distributed Systems, 4(8), 889–905. https://doi.org/10.1109/71.238624

Page | 5

You might also like