Introduction to performance testing

An Introduction to
Performance Testing
Richard Bishop
Trust IV Ltd

• This session aims to:
• Explain what performance testing is
….and why it is so important
• Explain performance test terminology
• Help you to understand the performance testing process
Objectives

• The process of executing a procedure or program or system with
the intent of finding errors.
• Should be part of all phases of SDLC.
What is Software Testing?

• Unit testing is the phase of software testing in which individual
units of source code are tested to determine whether they
are fit for use.
• Integration testing is the phase of software testing in
which individual software units are combined and
tested as a group.
• System testing of software or hardware is testing
conducted on a complete, integrated system to
evaluate the system's compliance with its specified
requirements.
Performance testing is appropriate in all of these test phases.
Just doing it at the end of the lifecycle causes problems.
Some Testing definitions

• Functional Testing
• Usually done first, to test application functionality not
performance.
• Very Comprehensive! Tries to cover all aspects of a
system/application.
• Tests at the Presentation Layer or user Interface (UI) level.
• Performance Testing
• Once the application passes functional testing, it is generally
submitted for performance testing
• During this process, the system/application is stressed to see how
well it performs when used by many users simultaneously
• Usually tests at the Protocol or Middleware level although more
modern techniques test at the UI.
Comparing Functional & Performance Testing

Comparing Functional & Performance Testing
Functional Testing Performance Testing
1
Tests the application’s
user interface, usability, and front-end
functionality
Does not test the application’s front-end
functionality
2
Does not test an application’s scalability or
resource usage
Tests an application’s scalability and monitors
resource usage
3
Can not determine how an application/system
will perform over time
Designed to determine how an
application/system will perform over time
4 Does not require a fully-functional application
Requires a fully-functional application for given
scenarios
5 One user Multiple users

• Quantification of risk
• Determine likelihood that application performance will meet the
business SLA’s
• Does not mitigate risk directly, but through identification and
quantification of risk, presents tuning opportunities
• Validation of application infrastructure
• Determine if the deployment environment is adequately sized for
the application to meet business SLA’s.
• Minimize extraneous hardware, software and the associated cost
of ownership
What is the Purpose of a Performance Test?

• How the application's response time changes if you increase
or decrease the number of users/transactions.
• How many users/transactions can simultaneously work with
the application without a perceptible slowdown.
• What load can crash the application.
• How hardware and software changes affect application
performance.
Performance Testing Helps to Determine

When Should You Consider Performance?
Production phase
What are the reasons
for degradations in
system performance? Deployment phase
Is the system reliable enough
to go into production?
Development phase
Does the system response
time meet SLA requirements?
Planning and Design phase
What is the best configuration to support
1000 users?Forecasting
How many users can
be added without
affecting system
performance?
At Every
Phase!

Why Consider Performance at Every Phase?
Because the cost of fixing software defects is much higher the later
they occur in the software development life cycle
£0
£1,000
£5,000
£125,000
Design Coding Unit Integration System Release to
production
Maintenance
£25,000

Why Performance Matters
0
1
2
3
4
5
6
7
8
9
1999 2003 2006 2009 2012
Acceptable Response Time (s)
“1 second” is the new “4 seconds”

• Validate the application to determine if it fulfills its performance goal or targets
• Identify the point at which the load becomes so great that the application fails
to meet performance requirements
• Determine if the application will support typical (normal) production load
conditions
• Tune the application for maximum performance by identifying the optimal
system configuration
• Reduce hardware / SaaS platform costs
• Locate bottlenecks including those in I/O, server, network, and database
Performance Testing Objectives

Application infrastructure
Web Servers
App Servers
DB Servers
External Users
Internal Users
Bad SQL
Slow Methods
Capacity limits
Contention Issues
WAN Sensitivity
Slow transactions
Performance Testing helps Identify
Configuration Issues

• Baseline
• Load
• Stress
• Soak
• Isolation
• Configuration
Basic types of Performance Test
http://en.wikipedia.org/wiki/Software_performance_testing

Types of performance test
0
20
40
60
80
100
120
10 30 50 70 90 110 130 150 170 190 210 230 250 270 290
vUsers
Elapsed time
“Normal” Load Profile
Ave load Peak load Typical test
0
20
40
60
80
100
120
10 30 50 70 90 110 130 150 170 190 210 230 250 270 290
vUsers
Elapsed time
“Soak Test” Load Profile
Ave load Peak load Soak test
“Normal” load profile
Common in performance tests
Soak test
Useful for identifying problems
which occur over time

Types of performance test
0
20
40
60
80
100
120
10 30 50 70 90 110 130 150 170 190 210 230 250 270 290
vUsers
Elapsed time
Load Profile
Ave load Peak load Spike test
0
50
100
150
200
250
300
350
10 30 50 70 90 110 130 150 170 190 210 230 250 270 290
vUsers
Elapsed time
Load Profile
Ave load Peak load Break test
“Stress/Break” test
Commonly requested
“Spike” test
Commonly requested

• 1 user for a set period of time or number of iterations
e.g. to determine pacing / think time
• A test at “normal load” to establish a baseline
for comparison with future tests
e.g. higher load or under different conditions
Baseline Test
Response time – 1 CPU 4GB RAM
Response time – 2 CPU 4GB RAM

• The “Classic” performance test
• Validates the system/application to determine if it meets the performance
targets.
• Provide information that will assist in performance tuning under various
workload conditions, hardware configurations, and database sizes
Load Test
• Helps identify key performance
data such as:
• Capacity
• Latency
• Response time
• Throughput
• Memory Management

SLA
Number of users
ResponseTime
Typically response time increase as virtual user numbers increase
1
10
20
25
10 50 100 200 500 1000 2000
5
15
30
35
40 Response Time
Load Test

• Identifies capacity limits of the system/application or infrastructure
• Identifies potential problems before going live
• Determines whether app can provide acceptable response times and
throughput under extreme conditions
Stress/Break Test
0
50
100
150
200
250
300
350
10 30 50 70 90 110 130 150 170 190 210 230 250 270 290
vUsers
Elapsed time
Load Profile
Ave load Peak load Break test

• Long Running test, identifies problems over time
– Test usually (not always) involves multiple business transactions
• Measures performance and efficiency of the system
(hardware and software) to:
– determine acceptable performance under load
– identify time-related problems such as memory leaks.
Soak Test

• Repeating a test which was known to cause a fault/defect.
• Used to “home in” on problems revealed by other forms of
performance testing. e.g. deadlocks, problems with specific test data
• May involve repetition of a load, stress or soak test scenario to point of
failure.
• Repeat test whilst changing code / application configuration until defect
can be closed.
Isolation Test

Used to determine optimum settings for configuration items such as:
• Load balancing
• Connection pool allocation
• Web, Mid-tier and Database configuration
• Common in “new build” projects
Configuration Test

Preparation
• Planning & requirements capture
• Determine realistic performance testing goals
• Prepare testing environment
• Prepare test data
• Identify performance testing use cases
• Capture, modify and validate use cases as scripts / scenarios
Performance Testing Process
Execution and Analysis
• Prepare environment / test data
• Execute performance test(s)
• Monitor impact of load on servers/database/networks
• Analyze results
• Submit defects
• System tuning
• Repeat execution and analysis process until
performance testing goals met.

It is nearly impossible to effectively performance test without automation!
• Manual testing is not manageable!
• Manual testing is not repeatable!
• Manual testing is not accurate!
Automated Testing Tools
Manual Automated


• Performance testing is very difficult without one
• Performance testing without a tool relies on hardware and personnel to
generate load, generally with wildly inaccurate results.
• Without automation the final test reports may be unreliable..
• User responses collected and compiled into a report may contain as
much human error as real data.
• The cost of repeating a 'manual' test is equal to cost of the first test.
When using a performance testing tool, costs go down as you repeat
your tests.
Benefits of an Automated Performance Testing Tool

• Most automated performance testing tools including LoadRunner
have the following components:
• Script Recorder (VuGen)
• Load Injector / Load Generator (Load Agents)
• Test Configuration and Management (Controller)
• Performance Monitor (Controller & other components)
• Analysis
How do Automated Performance Testing Tools Work?

Test Configuration
Client Network
AUT
HTTP
80/443
HTTP
80/443
Listener
443
8080
Controller
LG(s)
HTTP
80/443
LG(s)

Records end-user activity and generates transaction scripts.
Application Infrastructure
Web Servers
App Servers
DB Servers
End- User
HTTP – XML – SAP
Script Recorder

Generates a realistic application load, based on scripts and test parameters
Web Servers
App Servers
DB Servers
Generated Load
Load Generator aka. Injector

Load Generator
Load Generators can be local or remote
Options to place on AWS, Azure or HP Helion
Can also place on remote sites behind firewalls

• Mimics client behavior:
• Submits multiple requests to the AUT
• Waits for a period of time after the site sends a reply to the
request (think time) then submits a new request
• Can emulate thousands of concurrent users (virtual users) to test the
application scalability
How does a Load Injector work?
• Virtual User (VUs) mimic the behavior of a real user
• A performance test is valid only if a virtual users' behavior has
characteristics similar to those of actual users:
• Follow patterns similar to real users
• Use realistic think times

Test Configuration and Management
Test Controller manages the performance test and “controls” the action of load generators
Web Servers
App Servers
DB Servers
Generated Load
Test
Control

Capture server and network performance metrics
Web Servers
App Servers
DB Servers
Generated Load
Performance Monitor(s)

Post-test analysis of results
Web Servers
App Servers
DB Servers
Generated Load
Analyzing
Data
Analysis Module

Questions / Wrap up
• Want to have a go?
LoadRunner downloads and getting started guide
https://support.trustiv.co.uk/downloads/
• Want to learn more?
http://blog.trustiv.co.uk/
http://www.perfbytes.com/
http://alexanderpodelko.com/blog/
http://mtomlins.blogspot.co.uk/
http://www.developsense.com/
• Get in touch
@richardbishop
@TrustIV
richard.bishop@trustiv.co.uk
+44 (0) 7590 290646

Introduction to performance testing

Recommended

More Related Content

What's hot (20)

Similar to Introduction to performance testing (20)

Recently uploaded (20)

Introduction to performance testing