CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
Query optimization and challenges in DDBMS with Review Algorithms.
1. Pradip Raj Poudel (149-44), Kashiram
Pokharel(149-40)
“QUERY OPTIMIZATION
IN
DISTRIBUTEDDATABASE”
ME_CE III
NCIT, Lalitpur
A Review Article
By: Yasmeen Rm
Umar
Amit R Welekar
01/20/16
1
2. Outline:
Abstract
Introduction
Query Optimization
Optimization
Challenges
Steps In Query
Processing
S. Chaudhuri
Review
Fan/Xifeng Review
Chen/YU Review
Kossman/Stocker
Review
XUE Lin Review
Conclusion
First Part Second Part
01/20/16
2
3. Abstract:
Data is Growing over Distributed
Environment, Day By Day so Better
Distributed DBMS is Required.
Multiple sites with parts of Data’s ,so Query
optimization is a challenges in Distributed
Database.
Query optimization finds the best execution
plan from various options.
01/20/16
3
4. Introduction
All Data Placed on
Central Computer
location so Easy to
Access/Extract.
DB Query Easily
Transformed Into
RA operations.
No overhead
Data on multiple Sites
but centrally
Administrated.
Provides
Flexibility/customization
.
Ex. Location A can
Access data From
location B.
Location Transparency
Data Distributed, so
complex for Query
Transformation
Centralized Database Distributed Database
Database: Collection of
Files/Tables.
DBMS: Manage Database( CD or
DD)
01/20/16
4
5. Query
Optimization:
Data Distributed Over Different Sites in
Distributed Database.
If Query is Given, the response of that query
may Requires data From several Sites.
(DBMS fxn)
Now the Major task is “ Process A query with
location transparency and Find out Best
Sensible Execution Plan”.
Objective:
01/20/16
5
6. Optimization
Challenges:
1st
Break Query in Distributed Database
Environment.
2nd
Determine which Sites has less
Data/records.
As less Data ,less Communication and Vice-
versa.
Then Transfer those Data to Another Site.
More Sites= More Complex/Complication to
Process query.
Compute Cost using Effective Cost Module.
As Data Distributed in Different Sites, More Challenges To Compute Efficient Query
Plan.
01/20/16
6
7. Basic Steps In Query Processing
Plan
a). Query Decomposition:
Decompose into SimplerForm of RA.
OPTIMIZERCOMPONENTS:
a) . Query Engine
b) . Query Optimizer
b). Data localization:
Data Referenced to only one
location.(One Site)
c). Global Optimization:
Optimization of RA/Decision Making
Ex. Which site is efficient to move
data and where query will Execute.
d). Local Optimization:
When the Query Fragmented To
sites ,treat locally and Execute
Query.
01/20/16
7
8. Optimizer Components:
Query Engine:
a). Produce O/Pby taking I/P
and Performs Operations By
taking Physical
operators( Join,Sort,Loop).
b). Construct Parse tree which
shows flow of Data fromOne
Operation to AnotherOperation.
Query Optimizer:
a). Receives Parse Tree As I/P
From QE and Produce Best
Possible Execution Plan ,Based
On least Resource Consumption.
b). Not a Easy taskto generate
Efficient Query Plan
01/20/16
8
9. Review
Chaudhari Discussed on Basic Query
Optimization/Search Space/Cost Estimation
Technique.
Operator Tree having least resources
consumption would be best.
For Selecting Best plan, Statistical Info and
Execution cost Analyzed.
Statistical : No of Rows,memory,Joins,Pages
etc.
1. Surajit Chaudhari : Review
01/20/16
9
10. Review:
DD: Multiple Computer With Network.
GDBMS,LDBMS/CM are Elements of DB.
Distributed Database Manager is global and
local.
Proposed algorithm to improve semi-
connected sub query optimization to reduce
Network Cost.
But less efficient For Select Query.
2.Fan/XiFeng : Review
01/20/16
10
11. Review:
More Focused on Communication Cost.
Focused on Detail Study of Join/Semi join
Query.
The combination of Join & Semi join Results in
Large Reduction of Communication Cost.
Determines effect of join operation and find out
best combination of join which reduces
communication cost.
3.Chen/Yu: Review
01/20/16
11
12. Review:
Proposed Algorithm Based on IDP( iterative
Dynamic Programming)
Good But difficult to apply incase of Complex
queries.
Thus ,Uses Greedy Algorithm + DP concept
used For best Query plans.
Memory Requirements not Considered.
4.Kossmann/Stocker:Review
01/20/16
12
13. Review:
User Module: Analyze User Query
Syntax Analysis Module: done on Global Query
Query tree Conversion Module
Optimization Module: receives query tree which is optimized
and creates physical trees and calculates cost of each
physical operator tree.
Order Processing Module: Distribute Query to Server &
Returns result to user.
Local Data Dictionary used but table /cpu time/memory
increases.
5.XUE Lin: Review
01/20/16
13
14. Conclusion:
Dynamic Programming/Greedy: Large Space
Complexity.
Thus New Approach Used Based On Ant Colony
Algorithm, Where Each Relation is Considered as
Domain Value.
Better Execution Time has Been Achieved.
01/20/16
14