44
Philip Tellis @bluesmoon [email protected] SOASTA boomerang LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 1

Measuring the Performance Experienced by Real Users

Embed Size (px)

DESCRIPTION

My talk at the LSPE meetup in Jan 2013. http://www.meetup.com/SF-Bay-Area-Large-Scale-Production-Engineering/events/92123162/

Citation preview

Page 1: Measuring the Performance Experienced by Real Users

• Philip Tellis• @bluesmoon• [email protected]• SOASTA• boomerang

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 1

Page 2: Measuring the Performance Experienced by Real Users

Measuring the Performance Experienced byReal Users

Philip Tellis / [email protected]

LSPE Meetup / 2013-01-17

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 2

Page 3: Measuring the Performance Experienced by Real Users

Real Users

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 4: Measuring the Performance Experienced by Real Users

Real Users

• Get delighted• Get frustrated• Have credit cards• ... that they are attached to• Form an impression from experience

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 5: Measuring the Performance Experienced by Real Users

Real Users

• Get delighted• Get frustrated• Have credit cards• ... that they are attached to• Form an impression from experience

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 6: Measuring the Performance Experienced by Real Users

Real Users

• Get delighted• Get frustrated• Have credit cards• ... that they are attached to• Form an impression from experience

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 7: Measuring the Performance Experienced by Real Users

Real Users

• Get delighted• Get frustrated• Have credit cards• ... that they are attached to• Form an impression from experience

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 8: Measuring the Performance Experienced by Real Users

Real Users

• Get delighted• Get frustrated• Have credit cards• ... that they are attached to• Form an impression from experience

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 3

Page 9: Measuring the Performance Experienced by Real Users

What can we measure?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 4

Page 10: Measuring the Performance Experienced by Real Users

Anything that the user does!

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 5

Page 11: Measuring the Performance Experienced by Real Users

Anything that the user does!(in the browser)

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 5

Page 12: Measuring the Performance Experienced by Real Users

But we really care about performance here. . .

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 6

Page 13: Measuring the Performance Experienced by Real Users

So let’s talk about How

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 7

Page 14: Measuring the Performance Experienced by Real Users

Modern Browsers – no problem

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 8

Page 15: Measuring the Performance Experienced by Real Users

Modern Browsers – no problem

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 9

Page 16: Measuring the Performance Experienced by Real Users

Real Users don’t always use the latest browsers.

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 10

Page 17: Measuring the Performance Experienced by Real Users

What about older browsers?

1 Get a timestamp before they leave a page2 Save it in a cookie3 Compare with timestamp when the page loads

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 11

Page 18: Measuring the Performance Experienced by Real Users

What about older browsers?

1 Get a timestamp before they leave a page2 Save it in a cookie3 Compare with timestamp when the page loads

Then there’s some magic to handle edge-cases

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 11

Page 19: Measuring the Performance Experienced by Real Users

The magic. . . ...

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 12

Page 20: Measuring the Performance Experienced by Real Users

But it doesn’t need to be. . .

http://lognormal.github.com/boomerang/doc/

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 13

Page 21: Measuring the Performance Experienced by Real Users

boomerang

• BSD licensed• Measure load time, bandwidth, latency• DNS lookup time, NavTiming, Connection Type• Page Parameters• And anything else you can write a plugin for

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 14

Page 22: Measuring the Performance Experienced by Real Users

And it beacons all of this back to a predefined URL

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 15

Page 23: Measuring the Performance Experienced by Real Users

So what then?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 16

Page 24: Measuring the Performance Experienced by Real Users

Sparse Distribution

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 17

Page 25: Measuring the Performance Experienced by Real Users

Log-normal distribution

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 18

Page 26: Measuring the Performance Experienced by Real Users

Bimodal distribution

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 19

Page 27: Measuring the Performance Experienced by Real Users

Distributions

• Sparse distribution suggests that you don’t have enoughdata points

• Log-normal distribution is typical• Bi-modal distribution suggests two (or more) distributions

combined

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 20

Page 28: Measuring the Performance Experienced by Real Users

Distributions are interesting, but how does it affectengagement?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 21

Page 29: Measuring the Performance Experienced by Real Users

So we compared load time to bounce rate

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 22

Page 30: Measuring the Performance Experienced by Real Users

And we borrowed a biological term called LD50

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 23

Page 31: Measuring the Performance Experienced by Real Users

LD50

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 24

Page 32: Measuring the Performance Experienced by Real Users

This helps you fix only the stuff that’s hurting you.

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 25

Page 33: Measuring the Performance Experienced by Real Users

Did you know that IE7 seems to be faster than IE8?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 26

Page 34: Measuring the Performance Experienced by Real Users

But that’s mostly IE9 users browsing in IE7compatibility mode.

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 27

Page 35: Measuring the Performance Experienced by Real Users

Did you know that users bounce if a page is toofast?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 28

Page 36: Measuring the Performance Experienced by Real Users

But that’s mostly because error pages are the fastest.

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 29

Page 37: Measuring the Performance Experienced by Real Users

Did you know that user patience wrt performance isa cultural thing?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 30

Page 38: Measuring the Performance Experienced by Real Users

It’s tied both to where a user is from, and also thetype of service.

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 31

Page 39: Measuring the Performance Experienced by Real Users

Demo?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 32

Page 40: Measuring the Performance Experienced by Real Users

Summary

1 Measure the load time experienced by real users2 Correlate it with other things that user did3 Profit

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 33

Page 41: Measuring the Performance Experienced by Real Users

Summary

1 Measure the load time experienced by real users2 Correlate it with other things that user did3 Profit

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 33

Page 42: Measuring the Performance Experienced by Real Users

Summary

1 Measure the load time experienced by real users2 Correlate it with other things that user did3 Profit

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 33

Page 43: Measuring the Performance Experienced by Real Users

Thank You!Questions?

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 34

Page 44: Measuring the Performance Experienced by Real Users

• Philip Tellis• @bluesmoon• [email protected]• www.SOASTA.com• boomerang• LogNormal Blog

LSPE Meetup / 2013-01-17 Measuring the Performance Experienced by Real Users 35