Slide 1

Slide 1 text

Real  World  Problems,   Real  World  Load  Tes1ng   Jim  Holmes   Evangelist,  Test  Studio   @aJimHolmes   h7p://bit.ly/JimAtTestStudio   [email protected]  

Slide 2

Slide 2 text

Recent  News…  

Slide 3

Slide 3 text

What  the  CEO/customer  wants:   “Can  our  site  survive  as  we  grow?”     “Will  our  users  be  happy  when  the  system’s   being  used?”    

Slide 4

Slide 4 text

The  cost  of  degraded  performance   “According  to  a  2008  Aberdeen  Group  study,  a  1-­‐second  delay     in  page  load  Ome  equals  11  percent  fewer  page  views,  a  16     percent  decrease  in  customer  saOsfacOon,  and  a  7  percent     loss  in  conversions.  In  dollar  terms,  this  means  that  if  a  site     typically  earns  $100,000  a  day,  this  year  you  could  lose  $2.5     million  in  sales  for  every  second  slower  your  site  is  compared     to   last   year   (or,   for   that   ma7er,   compared   to   your   compeOtor’s  website).”    -­‐-­‐  Web  Load  Tes+ng  for  Dummies    

Slide 5

Slide 5 text

What’s  on  the  menu   •  Types  of  Load  TesOng   •  SeZng  ExpectaOons   •  What   •  Where   •  How   •  Demo!  

Slide 6

Slide 6 text

Types  of  Load  Tes1ng  

Slide 7

Slide 7 text

Soak  Tes1ng  

Slide 8

Slide 8 text

Failover  /  Tip  Tes1ng  

Slide 9

Slide 9 text

Load  Tes1ng  

Slide 10

Slide 10 text

Set  Expecta1ons  Early   •  What  are  our  realisOc  goals?   •  “Do  we  really  need  to  support  China  when  we  will   only  serve  our  neighborhood?”  -­‐-­‐  @BenCarey   •  Need  People,  Hardware,  TIME   •  What’s  on  the  table  for  fixing?  

Slide 11

Slide 11 text

WHERE  do  we  test?   •  Internal   •  Lets  us  focus  on  system  behavior   •  Hardware,  cache,  app  itself   •  External   •  Focus  is  on  overall  system,  usability  

Slide 12

Slide 12 text

WHAT  do  we  test?   •  ExisOng  site?  Learn  the  traffic  pa7erns.   •  How  do  users  use  your  site?   •  Got  history?  Hit  your  web  server  logs,  eg   •  Build  your  scenarios  to  mimic   •  Think  Omes   •  Ramp  up  

Slide 13

Slide 13 text

HOW  do  we  test?   •  Control  the  environment   •  Change  ONE  THING  at  a  Ome   •  Automate  setup  &  config   •  Baseline  data   •  System  config   •  Flush  caches,  app  pools,  etc   •  Warmup  scripts  

Slide 14

Slide 14 text

A  Sample  Load  Test   An   online   shopping   cart   applicaOon   is   required   to   support   10,000   concurrent   users   broken  out  into  following  acOviOes:     •  7500  Virtual  Users  (VU’s)  log  in,  browse  through  items  and  then  log  off.   •  1250  VU’s  log  in,  add  items  to  their  shopping  cart,  check  out  and  then  log  off.   •  750  VU’s  log  in,  request  RMA  and  then  log  off.   •  500  VU’s  just  log  in  without  any  subsequent  acOvity.   *Each  of  the  below  is  classified  as  a  ‘User  Profile’  or  ‘Scenario’.   -­‐Wikipedia  

Slide 15

Slide 15 text

How  We  Do  It.   Test  Studio   (create,  execute)   Controller  /   Reporter   Agent   (Generates  Load  /  VU’s)     Agent   (Generates  Load  /  VU’s)     Agent   (Generates  Load  /  VU’s)     ApplicaOon  

Slide 16

Slide 16 text

Virtual  User   Agent   (Generates  Load  /   VU’s)     ApplicaOon   HTTP(S)  Request  to  Server   HTTP(S)  Response  from  Server   HTTP(S)  Request  to  Server   HTTP(S)  Response  from  Server  

Slide 17

Slide 17 text

No content

Slide 18

Slide 18 text

Selec1ng  a  VU  Pack…  

Slide 19

Slide 19 text

Jim  Holmes   [email protected]   @aJimHolmes   @telerik  |  facebook.com/telerik