You are on page 1of 23

Why performance testing is

important ?

we offer the complete solution for
performance testing
Performance & Load Testing
Basics

Introduction to Performance Testing

Diference between Performance, Load and


Stress Testing

Why Performance Testing?

When is it required?

What should be tested?

Performance Testing Process

Load Test conguration for a web system

Practice !uestions
Introduction to Performance Testing
Performance testing is the "rocess of determining the
s"eed or efecti#eness of a com"uter, networ$,
software "rogram or de#ice%
&efore going into the details, we should understand
the factors that go#erns Performance testing'

Through"ut

(es"onse Time

Tuning

&enchmar$ing
Throughput
)a"ability of a "roduct
to handle multi"le
transactions in a gi#e
"eriod%
Through"ut re"resents
the number of
requests*business
transactions "rocessed
by the "roduct in a
s"ecied time duration%
+s the number of
concurrent users
increase, the
through"ut increases
almost linearly with the
number of requests% +s
there is #ery little
congestion within the
+""lication Ser#er
system queues%
Throughput
In the hea#y load ,one or Section &, as the concurrent client
load increases, through"ut remains relati#ely constant%
In Section ) -the buc$le ,one. one or more of the system
com"onents ha#e become e/hausted and through"ut starts
to degrade% 0or e/am"le, the system might enter the buc$le
,one when the networ$ connections at the Web ser#er
e/haust the limits of the networ$ ada"ter or if the requests
e/ceed o"erating system limits for le handles%
Response Time
It is equally im"ortant to
nd out how much time
each of the transactions
too$ to com"lete%
(es"onse time is
dened as the delay
between the "oint of
request and the rst
res"onse from the
"roduct%
The res"onse time
increases "ro"ortionally
to the user load%
Tuning
Tuning is the "rocedure
by which "roduct
"erformance is enhanced
by setting diferent
#alues to the "arameters
of the "roduct, o"erating
system and other
com"onents%
Tuning im"ro#es the
"roduct "erformance
without ha#ing to touch
the source code of the
"roduct%
Benchmarking
+ #ery well1im"ro#ed
"erformance of a
"roduct ma$es no
business sense if that
"erformance does
not match u" to the
com"etiti#e "roducts%
+ careful analysis is
needed to chal$ out
the list of
transactions to be
com"ared across
"roducts so that an
a""le1a""le
com"arison becomes
"ossible%
Performance Testing- Defnition
The testing to e#aluate the res"onse time -s"eed.,
through"ut and utili,ation of system to e/ecute its
required functions in com"arison with diferent #ersions
of the same "roduct or a diferent com"etiti#e "roduct is
called Performance Testing%
Performance testing is done to deri#e benchmar$
numbers for the system%
2ea#y load is not a""lied to the system
Tuning is "erformed until the system under test achie#es
the e/"ected le#els of "erformance%
Diference between
Performance Load and !tress Testing
Load Testing
Process of e/ercising the
system under test by feeding
it the largest tas$s it can
o"erate with%
)onstantly increasing the
load on the system #ia
automated tools to simulate
real time scenario with
#irtual users%
"#amp$es%
Testing a word "rocessor by
editing a #ery large
document%
0or Web +""lication load is
dened in terms of
concurrent users or 2TTP
connections%
Diference between
Performance Load and !tress
Testing
!tress Testing
Trying to brea$ the system
under test by o#erwhelming
its resources or by ta$ing
resources away from it%
Pur"ose is to ma$e sure that
the system fails and reco#ers
gracefully%
"#amp$e%
Double the baseline number
for concurrent users*2TTP
connections%
(andomly shut down and
restart "orts on the networ$
switches*routers that
connects ser#ers%
&h' Performance Testing(

The week of )eb * +,,,%


2ac$ers deli#ered o#er 31&illion transactions concurrently
to each one of these sites%

4ahoo

e&ay

&uy%com

+ma,on

eTrade
-ow did 'ou think the' performed (
&h' Performance Testing
Identies "roblems early on before they become costly to
resol#e%
(educes de#elo"ment cycles%
Produces better quality, more scalable code%
Pre#ents re#enue and credibility loss due to "oor Web site
"erformance%
5nables intelligent "lanning for future e/"ansion%
To ensure that the system meets "erformance e/"ectations
such as res"onse time, through"ut etc% under gi#en le#els of
load%
5/"ose bugs that do not surface in cursory testing, such as
memory management bugs, memory lea$s, bufer o#er6ows,
etc%
&hen is it re.uired(
Design Phase%
Pages containing lots of images and multimedia for
reasonable wait times% 2ea#y loads are less
im"ortant than $nowing which ty"es of content
cause slowdowns%
De/e$opment Phase%
To chec$ results of indi#idual "ages and "rocesses,
loo$ing for brea$ing "oints, unnecessary code and
bottlenec$s%
Dep$o'ment Phase%
To identify the minimum hardware and software
requirements for the a""lication%
&hat shou$d be tested(
-igh fre.uenc' transactions% The most frequently
used transactions ha#e the "otential to im"act the
"erformance of all of the other transactions if they are not
e7cient%
0ission 1ritica$ transactions% The more im"ortant
transactions that facilitate the core ob8ecti#es of the
system should be included, as failure under load of these
transactions has, by denition, the greatest im"act%
Read Transactions% +t least one (5+D 9:L4 transaction
should be included, so that "erformance of such
transactions can be diferentiated from other more
com"le/ transactions%
2pdate Transactions% +t least one u"date transaction
should be included so that "erformance of such
transactions can be diferentiated from other
transactions%
Performance Testing Process
Determine the performance testing ob3ecti/es
Describe the app$ication to test using a
app$ication mode$
3% Describe the 2ardware en#ironment
;% )reate a &enchmar$ -+genda. to be recorded in
Phase ;%
+% Dene what tas$s each user will "erform
&% Dene -or estimate. the "ercentage of users
"er tas$%
45P$anning
45P$anning
Record
(ecord the dened testing acti#ities that will be
used as a foundation for your load test scri"ts%
9ne acti#ity "er tas$ or multi"le acti#ities
de"ending on user tas$ denition
0odif'
<odify load test scri"ts dened by recorder to
re6ect more realistic Load test simulations%
Dening the "ro8ect, users
(andomi,e "arameters -Data, times, en#ironment.
(andomi,e user acti#ities that occur during the load
test
65 0odif'
65 0odif'
+5 Record
+5 Record
45 P$anning
45 P$anning

7irtua$ 2sers 872s9% Test :oa$s
Start' = <a/
(es"onse Time >? ;@ Sec
Incremented by' =
<a/imum' ;@@
Thin$ Time' = sec
Test !cript%
9ne ty"ical user from login through com"letion%
;5 "#ecute
;5 "#ecute
0onitoring the scenario% We monitor scenario e/ecution using
the #arious online runtime monitors%
<na$'sing test resu$ts% During scenario e/ecution, the tool
records the "erformance of the a""lication under diferent loads%
We use the gra"hs and re"orts to analyse the a""licationBs
"erformance%
*5 <na$'=e
*5 <na$'=e
>50onitor
>50onitor
Load Generation
System Under Test
Benefits of Automation
Web server Database
server
Vuser
host
Controller manages the virtual users
Analyze results with graphs and
reports
Replaces testers with virtual users
olves the resource limitations
Runs many Vusers on a few machines
Analysis
Controller
Too$s used for
Performance Testing
?pen !ource
9"enST+
Diesel Test
Test<a$er
Crinder
LoadSim
Dmeter
(ubis
1ommercia$
Load(unner
Sil$ Performer
!engine
5m"iri/ e1
Load

You might also like