Vous êtes sur la page 1sur 23

Performance Tuning of

Distributed Databases
Using Machine Learning
Techniques
Guide: Prof. S F Rodd
Presented by:
Project Group B10
1
Contents
Introduction
Need for tuning
Neural Network Approach to tuning
Memory Architecture of DB2 and
Oracle10g
Benchmark Factory Tool Overview
Distributed Database Environment
Distributed Tuning Architecture
GUI Snapshots
References


2
Introduction
Every enterprise uses DBMS to manage its day
to day functioning, forecast future growth and
develop strategies based on past data.

Wide spread use of Web Applications that are
database driven

Success of these apps. Depends on speed with
which the services are provided
3
Need for Tuning
Failure to provide the published web service causes huge
business losses.

Flash crash of NYSE in may 2010 wiped out $862billion in
just 20 minutes.

Manual tuning is expensive

According to payscale.com, a DBA earns on an avg anywhere
between Rs 1,69,242 - Rs 10,17,210

Changing/upgrading IT infrastructure every 2-3 years is
expensive and time consuming

Hence, there is a need to maximize the application of system
resources in an attempt to execute transactions as efficiently and
quickly as possible
4
Neural Network approach to Self-
Tuning
N

BHR

DBS
DB_CACHE, SH_POOL,
LRG_POOL
Training
Data
Set
DBMS
DATA
5
Neural Network based Tuning
Architecture

6
DB2 Memory Architecture
7
A Closer look at the Shared
Memory
8 Commands to alter tuning
parameters

Command to change the Buffer Pool Size
SQL> db2 -v alter bufferpool IBMDEFAULTBP size 4000
Command to check whether bufferpool has been altered
SQL>select * from syscat.bufferpools

BPNAME BUFFERPOOLID KSIZE NGNAME NPAGES PAGESIZE ESTORE
---------- ----------- ------------ ------ --------- -------- -----------
IBMDEFAULTBP 1 0 - 4000 4096 N

To change the LOCKHEAP parameters, use the following commands:
SQL> db2 -v update db cfg for DB_NAME using LOCKLIST a_number
SQL> db2 -v update db cfg for DB_NAME using MAXLOCKS b_number
SQL> db2 -v update db cfg for DB_NAME using LOCKTIMEOUT c_number

To change the values of SORTHEAP
SQL> db2 -v update db cfg for DB_NAME using SORTHEAP a_value

9
Oracle10g Memory Architecture
10
Auto-tuning features in Existing
DBMS
Auto-tuning feature is available in Oracle 10g
SQL> show parameter sga_target;
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
sga_target big integer 0
SQL> show parameter sga_max_size;
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
sga_max_size big integer 164M
SQL> alter system set sga_max_size=1200M scope=spfile;
SQL> alter system set sga_target=1100M scope=both;
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
sga_target big integer 1104M

11
Benchmark Factory Tool


12
Distributed Database
Environment
Types
Homogeneous
Heterogeneous
Applications accessing data across databases
using Database links
Link creation:

CREATE DATABASE LINK foo CONNECT TO scott
IDENTIFIED BY tiger USING 'rev';

SELECT * FROM emp@foo
13
Distributed Tuning Architecture
14
GUI Snapshots
15
16
17
Response Time Graph For Oracle
10g
18
Response Time Graph For DB2

0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0 10 20 30 40 50 60 70
Avg. Response Time (sec)
Avg. Response Time (sec)
19
CONCLUSION

The proposed self tuning architecture enhances the
performance of the DBMS
On distributed databases in Oracle 10G our approach
yields a significant improvement in performance as
compared to in-built Auto-tuning feature of Oracle.
NN-based tuning approach on centralized DB2
database yields a significant improvement in
performance as compared to self-tuning feature of
DB2.
NN-based tuning approach is a generic one and can
be used to tune any DBMS.

20
Future Plan
Implement NN tuning on DB2 on standalone
server and extend the same to Distributed
environment
Test on other DBMS like MSSQL Server,
MySQL to validate the method and make it
more generic approach.
Develop a self-tuning approach based on other
machine learning techniques like statistical
approach.


21
References
[1]Fuzzy Controlled Architecture for PerformanceTuning of Database Management System by SF Rodd ,
Umakant P Kulkarni, A.R. Yardi

[2]Enhanced Performance of Database by Automated Self-Tuned Systems by Ankit Verma.

[3]Adaptive Self-Tuning Memory in DB2 by Adam J. Storm, Christian Garcia-Arellano,
Sam S. Lightstone, Yixin Diao , M. Surendra .

[4]A New Approach to Dynamic Self-Tuning of Database Buffers by DINH NGUYEN TRAN, PHUNG
CHINH HUYNH, Y. C. TAY and ANTHONY K. H. TUNG.

[5]AutoAdmin: Self-Tuning Database Systems Technology by Sanjay Agrawal, Nicolas Bruno, Surajit
Chaudhuri, Vivek Narasayya

[6] Sanjay Agarwal, Nicolas Bruno, Surajit Chaudhari, AutoAdmin: Self Tuning Database System
Technology, IEEE Data Engineering Bulletin, 2006.

[7] Wiese, David; Rabinovitch, Gennadi, Knowledge Management in Autonomic Database Performance
Tuning, 20-25 April 2009.



22
THANK YOU
23

Vous aimerez peut-être aussi