velocity eu 2013 what is the velocity of an unladen swallow?
DESCRIPTION
Seatwave was growing fast, success was unabated, and industry awards were landing on their doormat. Infrastructure had been revamped, load patterns were understood. Everything was going just great… Until… The marketing team planned Seatwave’s first UK TV campaign – all regions – simultaneously, but only told the engineering team the day before the first advert was due to run! 10 seconds into the advert the site melted and there was a collective thud as heads hit desks. It was expensive lesson to learn but also the wake up call that forced everyone in Seatwave to focus on the performance of their site. In this session we’ll share that pain we experienced, and how we improved performance so that when all our competitors crashed during the UKs largest concert ticket sale, we were able to take 20 days revenue in just 2 hours! However, maintaining performance is a challenge, product owners want new features, the site starts to put on weight and slowly performance starts to degrade once more. Will it take another disaster to focus everyone on performance or is there another way to avoid “boom and bust”? We’ll talk about the steps we’re taking to avoid “boom and bust” by making both performance and the impact performance has on our customers visible to everyone across Seatwave including: Our Adobe Site Catalyst installation with a custom implementation of the W3C Navigation Timing API allowing us to segment our business KPI’s by speed. How we’re using a WebPageTest within continuous integration for our QA and production builds. How we constantly review our performance against competitors using our own installation of the HTTPArchive. Join us on our quest in search of the Holy Grail of truly understanding how web site performance affects our business, and the processes and systems we are putting in place to ensure we keep speed at the heart of our product development roadmap.TRANSCRIPT
http://www.flickr.com/photos/damien_m_in_japan/575938779/
http://www.flickr.com/photos/gareth1953/7057132797
http://www.flickr.com/photos/radulovski/7837486702/
http://www.flickr.com/photos/yimhafiz/5119955475/
http://www.flickr.com/photos/evablue/6184132023
http://www.flickr.com/photos/ful1to/3783198574/
http://www.flickr.com/photos/imnohero/2330548144
Photo courtesy of Florian Wieder wiederdesign.com
http://www.flickr.com/photos/24365773@N03/9211817009
http://www.flickr.com/photos/julianlim/4598412264/
http://www.flickr.com/photos/8001265@N08/2259250324
http://www.flickr.com/photos/capttaco/6939970676
http://www.flickr.com/photos/westmidlandspolice/8169072160/
http://www.flickr.com/photos/neospire/3595638270
0
10
20
30
40
50
60
70
80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
30
/07
/20
10
06
/08
/20
10
13
/08
/20
10
20
/08
/20
10
27
/08
/20
10
03
/09
/20
10
10
/09
/20
10
17
/09
/20
10
24
/09
/20
10
01
/10
/20
10
08
/10
/20
10
15
/10
/20
10
22
/10
/20
10
29
/10
/20
10
DB
Cal
ls P
er
Sess
ion
CP
U P
er
Sess
ion
(se
c)
Database Usage per Web Session
CPUPerSession
DBCallsPerSession
http://www.flickr.com/photos/olly247/2831803988
Queuing
iRule
Data Group
Site | Limit
Main Site
Site Limit Checks
When client
timer reaches
zero
User
All Limit Checks
Does
Site Limit
Exist?
All Sites Limit
Exceeded?
Does
All Sites Limit
Exist?
Site Limit
Exceeded?
Does
Cookie
Exist?
Traffic
counters Increment /
Decrement
No NoYes
Yes Yes YesNo
Yes
No No
Server
controlled timer
served to client
Infr
as
tru
ctu
re A
Infr
as
tru
ctu
re B
Original client request is
passed from queue back
to F5 so customer goes to
the exact page they
originally requested when
queue is inactive
29-Oct-2010
One site didn’t go down….
0
10
20
30
40
50
60
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
Traf
fic
mu
ltip
le
Time
Traffic Profile 29th October 2010
1 months revenue in 1 day
10 days revenue per hour for 2 hours
http://www.flickr.com/photos/ful1to/3783198574/
http://www.flickr.com/photos/katescars/5228508617
http://www.flickr.com/photos/imnohero/2330548144
Photo courtesy of Florian Wieder wiederdesign.com
http://www.flickr.com/photos/bensutherland/5920185547
0
10
20
30
40
50
60
19
:00
20
:00
21
:00
22
:00
23
:00
Traf
fic
mu
ltip
le
Time
Traffic Profile 14th November 2010
Spikes within 15 seconds of TV advert start
Traffic boost in ad breaks when we had no active advertisement
http://www.flickr.com/photos/75001512@N00/3938175748
http://www.flickr.com/photos/epsos/4929687589
complacency
Com·pla·cen·cy[kuhm-pley-suhn-see]
a feeling of quiet pleasure or security, often while unaware of some potential danger, defect, or the like;
http://www.flickr.com/photos/pointshoot/861934878
http://www.flickr.com/photos/zeevveez/7095563439
http://www.flickr.com/photos/tico24/57164061
http://www.flickr.com/photos/bluefountainmedia/8191245592
2013 : Analytics Refocused
http://www.flickr.com/photos/sebilden/8472188590
http://www.meetup.com/London-Web-Performance-Group/events/100061982/ http://www.webwob.com/html/webpagetest.html
http://www.flickr.com/photos/dan_culleton/4908793028
WebPageTest Private Instance • https://sites.google.com/a/webpagetest.org/docs/private-instances • http://andydavies.me/blog/2012/09/18/how-to-create-an-all-in-one-webpagetest-private-instance/ HTTPArchive Private Instance • We will publish the HTTPArchive customization we have developed – watch this space! WebPageTest / Jenkins mash-up • http://www.webwob.com/html/webpagetest.html Events • http://velocityconf.com/velocityeu2013/public/schedule/detail/32943 • http://www.webperfdays.org/events/2013-london/index.html • http://www.meetup.com/London-Web-Performance-Group/
http://www.flickr.com/photos/slgc/5970360335
http://www.seatwave.com/movember
http://www.flickr.com/photos/damien_m_in_japan/575938779/