Envision , Evolve, Unleash
Performance Testing - LR Venu Gopal Reddy.S
Contents • • • • • •
Why Load Test Your Web Application ? Functional vs. Load Web Testing Web-Based, Multi-Tiered Architecture Manual Testing Is Problematic The LoadRunner Solution LoadRunner Expert Workflow “The Big Picture” • Plan Load Test • Sample Task Distribution Diagram • Transaction Profile
10/17/08
2
Contents Contd… • • • • • • • • •
User Profile How to Use the Information Create VuserScript Actions Think Times Parameterization Determine Which Fields to Parameterize Parameterization Decision-Maker Select the Data Access Method
10/17/08
3
Contents Contd… • • • • • • •
Data File Replacement Methods Run-Time Settings – Iterations Summary - Parameterization Transactions Automatic Transactions Summary Next session……….
10/17/08
4
Why Load Test Your Web Application?
• Why Load Test Your Web Application? – The failure of a mission-critical web application can be costly • don’t just cross your fingers
– Deploy with confidence • assure performance and functionality under realworld conditions
10/17/08
5
Functional vs. Load Web Testing Functional test
Load test
OBJECTIVE
EXAMPLE
Functionality
Do business processes function properly after implementation?
OBJECTIVE
EXAMPLE
Stability
Will 2,000 concurrent hits crash the server?
Performance
Is response time acceptable according to specifications?
Functionality under load
Do business processes function properly under heavy load?
10/17/08
6
Web-Based, Multi-Tiered Architecture
10/17/08
7
Manual Testing Is Problematic Do you have the testing resources? • Testing personnel • Client machines
All of you, click the GO button again
How do you synchronize users? How do you collect and analyze results? How do you achieve test repeatability? Coordinator
Analysis?
Web server
Database server
123.20
Testers
Load Generation
System Under Test 10/17/08
8
The LoadRunner Solution Overcomes resource limitations
Controller
Analysis
• Replaces testers with “Virtual Users” • Runs many Vusers on few machines • Controller manages the Vusers • Meaningful results with analysis tools • Repeats tests with scripted actions
Vuser host
Web server
Load Generation
Database server
System Under Test 10/17/08
9
The LoadRunner Solution Provides support for many protocols and APIs
10/17/08
10
The LoadRunner Solution
Virtual User Generator
Creates the Vusers deployed by the Controller
LoadRunner Controller
Generates load and collects test results
LoadRunner Analysis
Compiles and displays test results with graphical and statistical tools
10/17/08
11
LoadRunner Expert Workflow “The Big Picture” Phase 1
Phase 2
Phase 3
Phase 4
Phase 5
Plan Load Test
Create Web Virtual Users
Create Scenarios
Run Scenarios
Tune Analyze System Under System Based on Analysis Load
LoadRunner VUGEN
LoadRunner
CONTROLLER& ANALYSIS
10/17/08
12
Plan Load Test • System Usage Information – Task Distribution Diagram • which business tasks? • how many operations at what times of the day?
– Transaction Profile • how many operations on average, how many at peaks? • how much database activity? • how much risk to business if task fails?
– User Profile • which tasks does each real user perform? • what is ratio of different tasks for each user?
10/17/08
13
Sample Task Distribution Diagram Hours to be emulated
Business Processes Logout
40
Night
30
Day40
75
Night
Compose
1000
1200
Inbox
180
210
150
130
100
150
200
170
140
Login
*Yahoo
50
50
*System backup
1
1
12 am
2
600
50
4
50
6
8
10
12 pm
2
4
6
8
10
12
ONLINE BOOKSTORE (typical day, thousands of users) *not a web task
10/17/08
14
Transaction Profile Which business process to automate? Business process name
Typical day Peak day
Web server activity
Databas e activity
Risk
Heavy
Light
High
✓
70/hr
✓ ✓
10/hr
15/hr
Moderate
Moderate
Low
130/hr
✓ ✓ 180/hr
Moderate
Moderate
Mod.
New Folder
20/hr
30/hr
Moderate
Moderate
High
Logout
40/hr
90/hr
Moderate
Heavy
High
Log in Inbox Compose
Selection criteria:
210/hr
✓
✓
✓ ✓
✓ High throughput ✓ High in database I/O ✓ High business risk (should business process fail) 10/17/08
15
User Profile
Business Process
Folder 1
Folder 2
No. of Concurrent Users
15
5
InBox
10/hr
5/hr
Login
5/hr
25/hr
Logout
Folder 3 10
5/hr
Compose
5/hr
10/17/08
16
How to Use the Information • Decide which business processes to automate • Learn what processes are linked (data dependent) • Discover peak load times – peak hours of the day – peak days of the month, quarter, or year • Decide whether to emulate processes or users
10/17/08
17
Create VuserScript •Record Vuser Actions • Add Load Runner Transactions •Parameterize data • Verify correct Execution
10/17/08
18
Actions • A recorded business process that, when played back, emulates a real user performing the business process actions on a system. Logon (recording optional)
vuser_init.c
User Actions
Logoff
(Business Processes)
(recording optional)
Action1.c, Action2.c, etc. (e.g., Create Order, Ship Order)
vuser_end.c
This section may be iterated (repeated) during one test run
Virtual User Playback Model
10/17/08
19
Think Times
Think time steps: • Simulate human users by pausing between user actions •
•
Inserted automatically by VuGen during recording Default Run-Time Setting is “Ignore think times” *(VuGen only)
10/17/08
20
Parameterization • The practice of replacing a recorded value in a Vuser with a placeholder which can represent an expandable range of values. • Solve certain problems that may occur during playback – date constraints – unique constraints – data caching – data dependency • Emulate real user activity • Exercise the server
10/17/08
21
Determine Which Fields to Parameterize
• Which fields require a current date? • Which fields require unique values? • Which fields exercise the database server? • Which fields’ values are dependent for validity on the value of another field?
10/17/08
22
Parameterization Decision-Maker For each field, ask . . .
Is there a unique constraint?
Do not parameterize No
Is there a date constraint?
No
Does data get cached?
Yes Yes
No No
Is this a data dependent field?
Yes Yes
Parameterize
Parameterize 10/17/08
Parameterize
Parameterize 23
Select the Data Access Method
• • • •
10/17/08
Sequential Random Unique Same line as ...
24
Data File Replacement Methods Sequential Vuser Iteration
1 2 3 4
Random
Unique
Vuser 1
Vuser 2
Vuser 3
Vuser 4
a a b c c d d h
a f b e c j d g
a l b m c k d n
a i b c c e d b
a b c d
e f g h
10/17/08
i j k l
m n param1
o p 25
Run-Time Settings – Iterations For the Sequential access method, Iteration Count should equal the number of rows in the data file
10/17/08
26
Summary - Parameterization • Parameterize fields in order to: – solve unique data constraints – solve date constraints – exercise the server(s) – handle data dependency • Establish input data: – from the database – with the help of a Functional Expert • Control Vuser execution from the Run-Time Settings – extended log, parameter substitution – number of iterations 10/17/08
27
Transactions • An end-to-end (browser-to backend-to browser) measurement of one or more user actions within an action file.
Why Add LoadRunner Transactions? • Transactions are LoadRunner’s only means of measuring performance • Measure high business risk transactions • Able to measure a single step or a group of steps • Allows performance comparison between different load tests • Helps to identify performance problems 10/17/08
28
Transactions Cont…. • Start and End Transaction
• During recording Start and End Transaction • After recording Start and End Transaction • LR_AUTO – use this when you want LoadRunner to detect the transaction status automatically • LR_PASS / LR_FAIL – use one of these when you want to set the transaction status manually in the code
10/17/08
29
Automatic Transactions • Measure performance of every test step • Compare individual steps to each other • Compare individual steps to entire business process Steps 12 sec 1 sec 9 sec 12 sec 1 sec 21 sec 19 sec 4 sec
10/17/08
Complete business process (1 minute, 19 seconds total)
30
Summary • Use LoadRunner transactions to measure: – end-to-end response times of specific steps in the business process – the entire business process – every step in the business process automatically • Add LoadRunner transactions during or after recording • Use automatic LoadRunner transactions to measure every step without manually inserting transactions
10/17/08
31
Next session……….
• Controller • Analysis • Advanced topics
10/17/08
32
Thank You
10/17/08
33
Goal Oriented Approach Phase wise Approach Goal definition (g) Preparation (p) Test execution (e) Reporting + decision next phase (rd) Multiple iterations of phase (p), (e) and (rd) are planned at a maximum of 3. In general the more important a certain activity is the earlier it is scheduled (i.e. cycle 1). The overall project process can be depicted as follows: g Goal definition
p1,2,3 Test prep.
e1,2,3 Test execution
rd1,2,3 Test reporting
Same cycle
Decision
Finished!
Next cycle Max. 3 cycles
10/17/08
34
Architecture
10/17/08
35