web performance optimisation

Post on 15-May-2015

589 Views

Category:

Technology

0 Downloads

Preview:

Click to see full reader

DESCRIPTION

The Need for Speed, a presentation at the Melbourne SEO Meetup on "Web Performance Optimisation and Why It Matters".

TRANSCRIPT

I  feel  the  need…  the  need  for  SPEED!  

@chrisburgess                                                                                                                                                                              ChrisBurgess.com.au    

Web  performance  op8misa8on  and  why  it  ma=ers  

We  know  page  speed  is  a  ranking  factor  for  search  engines  

November  13,  2009  

Google  first  started  talking  about  speed  4  years  ago  

Today  

So  we  get  that  Google        thinks  it’s  a  big  deal…  

But  remember,  we’re  not  just  op8mising  for  search  engines,  page  speed  is    

a  huge  factor  for  users  

Data  from  h5p://www.strangeloopnetworks.com/assets/images/Infographics/Web-­‐Stress-­‐Infographic-­‐500.jpg  Image:  h5p://fisEuloFalent.com/wp-­‐content/uploads/2012/06/angry-­‐computer-­‐large-­‐500x320.jpg  

78%  of  users  say  they’ve  felt  stress  or  anger  while  using  a  slow  website  

Data  from  h5p://www.strangeloopnetworks.com/assets/images/Infographics/Web-­‐Stress-­‐Infographic-­‐500.jpg  Image:  h5p://celebrity-­‐lists.com/wp-­‐content/uploads/2012/03/brad-­‐pi5.jpg  

4%  of  people  have  thrown  their  phone  while  using  a  slow  mobile  site  

47%  of  consumers  expect    a  web  page  to  load  in  2  seconds  or  less  

Data  from  h5p://blog.kissmetrics.com/loading-­‐Nme/?wide=1  Image  from:  h5p://firewalkercreaNve.com/wp-­‐content/uploads/2011/01/2seconds.jpg  

Data  from  h5p://blog.kissmetrics.com/loading-­‐Nme/?wide=1    

Page  load  )me  in  seconds  

Page  aba

ndon

ment  %

 

25%  

50%  

2   4   6   8   10  

Source:  h5p://www.strangeloopnetworks.com/assets/images/Infographics/Web-­‐Stress-­‐Infographic-­‐500.jpg  

Why  do  happy  users  ma=er?  

Source:  h5p://www.strangeloopnetworks.com/assets/images/Infographics/Web-­‐Stress-­‐Infographic-­‐500.jpg  

1  second  delay  

7%    reduc8on  

 in  conversions  

Data  from  h5p://blog.kissmetrics.com/loading-­‐Nme/?wide=1  

Data  from  h5p://blog.kissmetrics.com/loading-­‐Nme/?wide=1  Image:  h5p://workforceplanningtools.com.au/wp-­‐content/uploads/2013/01/how-­‐to-­‐lose-­‐money.jpg      

If  your  site  made  $100K  a  year  that’s  $7K  in  lost  revenue  for  1  second!  

So  happy  users  are  more  likely  to:    ▷  Hang  around  longer  ▷  Come  back  to  your  site  ▷  Become  engaged  ▷  Convert  ▷  Talk  about  your  site  ▷  Share  your  content  ▷  Link  to  your  site  

“We  want  you  to  be  able  to  flick  from  one  page  to  another  as  quickly  as  you  can  flick  a  page  in  a  book.  So  we’re  aiming  very  

very  high  …  at  something  like  100  milliseconds”    –  Urs  Hölzle,  Senior  VP  Opera5ons,  Google  

Source:  h5p://www.strangeloopnetworks.com/assets/images/Infographics/Web-­‐Stress-­‐Infographic-­‐500.jpg      Image:  h5p://images.huffingtonpost.com/gen/143651/FLIP-­‐BOOK-­‐RAINBOW.jpg  

Source:  h5p://www.strangeloopnetworks.com/assets/images/visualizing_web_performance_poster.jpg-­‐RAINBOW.jpg  

In  2010,  the  Mozilla  team  found  the  Firefox  download  page  was  loading  in  7  seconds.    

 …by  reducing  the  average  page  load  8me  by  2.2  seconds,  they  saw  a  15.4%  increase  in  downloads.  

Source:  h5p://zoompf.com/blog/2013/08/web-­‐performance-­‐basics-­‐for-­‐the-­‐markeNng-­‐team  

How  fast  is  fast  enough?    ▷  Within  the  user  interface  (UI)  ▷  0.1  second  =  User  feels  that  the  system  is  reacNng  

instantaneously    ▷  1  second  =  Limit  for  the  user's  flow  of  thought  to  stay  

uninterrupted  ▷  10  seconds  =  Limit  for  keeping  the  user's  a5enNon  focused,  any  

longer  users  will  want  to  perform  other  tasks  while  waiNng,  so  they  should  be  given  feedback  indicaNng  when  the  computer  expects  to  be  done.  

Source:  h5p://www.nngroup.com/arNcles/response-­‐Nmes-­‐3-­‐important-­‐limits/  

The  appearance  of  performance…  

Preloading  vs  Lazy  Loading  

▷  Preloading  ▷  Lazy  Loading  

So  how  do  we  make  our  sites  faster?  

Client  side  

Server  side  

80  /  20  Rule  

There  is  no  one  size  fits  all…  

You  must  test,  then  test  some  more…  

Performance  Tes8ng  vs  Load  Tes8ng  

There  are  six  fundamental  reasons  companies  measure  performance  of  their  sites:    ▷  Establish  baselines  ▷  Detect  and  repair  errors  ▷  Measure  the  effecNveness  of  change  ▷  Determine  the  impact  of  an  outage  ▷  Resolve  disputes  with  users  ▷  EsNmate  how  much  capacity  will  be  needed  in  

the  future  

Source:  “Complete  Web  Monitoring”  Alistair  Croll  and  Sean  Power  

Google  Best  Prac8ces  h5ps://developers.google.com/speed/docs/best-­‐pracNces/rules_intro  

 Yahoo!  Performance  Rules  

h5p://developer.yahoo.com/performance/rules.html    

…and  there’s  new  content  being  added  all  the  )me.    

But  connec8ons  are  gefng  faster,  right?  

Source:  h5p://h5parchive.org/  

Source:  h5p://h5parchive.org/  

Source:  h5p://h5parchive.org/  

Processing  generaNon-­‐generosity…  —  [original]  300  x  178:  Reduced  by  8.9%  (5.3  KB)  —  [thumbnail]  150  x  150:  Reduced  by  11.5%  (3.1  KB)  —  [medium]  300  x  178:  Reduced  by  8.9%  (5.3  KB)    Processing  optus_rockcorps_infographic_print…  —  [original]  600  x  1286:  Reduced  by  8.0%  (15.2  KB)  —  [thumbnail]  150  x  150:  Reduced  by  6.1%  (596  B)  —  [medium]  140  x  300:  Reduced  by  6.5%  (1.1  KB)  —  [large]  477  x  1024:  Reduced  by  7.6%  (10.2  KB)    Processing  australian-­‐social-­‐media-­‐staNsNcs-­‐2012-­‐vs-­‐2013_small…  —  [original]  600  x  450:  Reduced  by  11.4%  (7.8  KB)  —  [thumbnail]  150  x  150:  Reduced  by  8.1%  (666  B)  —  [medium]  300  x  225:  Reduced  by  8.7%  (1.7  KB)    Processing  australian-­‐social-­‐media-­‐staNsNcs-­‐2012-­‐vs-­‐2013_large…  —  [original]  1020  x  765:  Reduced  by  14.8%  (21.6  KB)  —  [thumbnail]  150  x  150:  Reduced  by  7.5%  (633  B)  —  [medium]  300  x  225:  Reduced  by  8.2%  (1.6  KB)  

Also  check  out…    

▷  h5p://compresspng.com/  

▷  h5p://pnggauntlet.com/  

▷  h5p://imageopNm.com/  

What  can  make  your  site  slow?    ▷  HosNng  ▷  Use  of  media  (e.g  photos  and  video)  ▷  CSS,  JavaScript,  Fonts  ▷  PlaEorm  (theme,  plugins)  ▷  Errors  and  misconfiguraNon  ▷  SSL  (eek!)  

There’s  a  plugin  for  that!    ▷  WP  Minify  h5p://wordpress.org/plugins/wp-­‐minify/  

▷  WP  Smush.it  h5p://wordpress.org/plugins/wp-­‐smushit/  

▷  WP  Super  Cache  h5p://wordpress.org/plugins/wp-­‐super-­‐cache/  (but  there  are  others)  

▷  Plugin  Organizer  h5p://wordpress.org/plugins/plugin-­‐organizer/  

▷  JS  &  CSS  Script  OpNmizer  h5p://wordpress.org/plugins/js-­‐css-­‐script-­‐opNmizer/  

▷  WP-­‐DBManager  h5p://wordpress.org/plugins/wp-­‐dbmanager/  

▷  P3  (Plugin  Performance  Profiler)  h5p://wordpress.org/plugins/p3-­‐profiler/  (see  next  slide)  

Source:  “Complete  Web  Monitoring”  Alistair  Croll  and  Sean  Power  

P3  Profiler  

▷  h5p://wordpress.org/plugins/p3-­‐profiler/  

Look  for:    

#  ##############################################################################  #                                                                                                                                              WEB  PERFORMANCE                                                                                                                        #  #  ##############################################################################  

 

h5ps://github.com/h5bp/html5-­‐boilerplate  

Does  everybody  know  what  8me  it  is?  

Tool  8me  -­‐  tools  for  tes8ng  

▷  WebPageTest.org  

▷  Google  AnalyNcs  ▷  Yslow  ▷  Your  Web  Brower  “Developer  Tools”  

▷  Pingdom  

▷  LoadImpact.com  

WebPageTest.org  (Example)  

Google  Analy8cs  

YSlow  

PageSpeed  Extension  

Also  check  out  Speed  Tracer  for  web  apps  

Web  Browser  (Developer  Tools)  

Google  Tools  ▷  h5ps://developers.google.com/speed/pagespeed/  

Yo=aa  ▷  h5p://www.yo5aa.com/  

•  h5p://loadimpact.com/  

Load  Impact  

▷  h5p://www.200Please.com  

200Please.com  

▷  h5ps://www.pingdom.com/  

Pingdom  

More  tools…  

▷  ab  ▷  cURL  ▷  cURL-­‐loader  ▷  h5perf  ▷  jmeter  

▷  Siege  

ab  -­‐n  100  -­‐c  10  h=p://example.com/  

100  HTTP  GET  requests,  10  requests  at  a  Nme  

cmd:~  cb$  ab  -­‐n  100  -­‐c  10  h=p://test.hypercrao.com/  This  is  ApacheBench,  Version  2.3  <$Revision:  655654  $>  Copyright  1996  Adam  Twiss,  Zeus  Technology  Ltd,  h=p://www.zeustech.net/  Licensed  to  The  Apache  Sooware  Founda8on,  h=p://www.apache.org/    Benchmarking  test.hypercrao.com  (be  pa8ent).....done      Server  Sooware:                Apache/2.2.11  Server  Hostname:                test.hypercrao.com  Server  Port:                        80    Document  Path:                    /  Document  Length:                22595  bytes    Concurrency  Level:            10  Time  taken  for  tests:      8.553  seconds  Complete  requests:            100  Failed  requests:                0  Write  errors:                      0  Total  transferred:            2303528  bytes  HTML  transferred:              2281598  bytes  Requests  per  second:        11.69  [#/sec]  (mean)  Time  per  request:              855.335  [ms]  (mean)  Time  per  request:              85.533  [ms]  (mean,  across  all  concurrent  requests)  Transfer  rate:                    263.00  [Kbytes/sec]  received    Connec8on  Times  (ms)                              min    mean[+/-­‐sd]  median      max  Connect:              30      33      2.0          33            39  Processing:      350    764  281.9        707        1495  Wai8ng:            220    474  233.2        388        1115  Total:                382    797  282.7        739        1531    Percentage  of  the  requests  served  within  a  certain  8me  (ms)      50%        739      66%        802      75%        876      80%        995      90%      1333      95%      1440      98%      1461      99%      1531    100%      1531  (longest  request)  

Other  considera8ons  

Mobile  

High  defini8on  displays  

Social  

Where  to  get  more  informa8on  

Steve  Souders  

Source:  h5p://stevesouders.com/  

Planet  Performance  

Source:  h5p://www.perfplanet.com/  

Source:  h5p://moz.com/ugc/why-­‐site-­‐speed-­‐opNmisaNon-­‐should-­‐be-­‐part-­‐of-­‐your-­‐seo-­‐strategy  

Moz  

Recap    So  we’ve  talked  about:  

▷  Why  speed  is  important  and  how  it  can  help  your  site  

▷  Why  it’s  important  to  keep  users  happy  

▷  Sweet  spot  =  aim  for  1  second  or  less  

▷  Loading  cues  

▷  TesNng  

▷  Tools  

▷  Where  to  get  more  informaNon  

Source:  “Complete  Web  Monitoring”  Alistair  Croll  and  Sean  Power  

In  summary…  

Search  engines  care  about  speed…  

Users  care  about  speed…  

Search  engines  care  about  users…  

 So  you  should  feel  the  need…the  need  for  speed!  

 

Thank  you  

top related