An Overview of Amazon Silk Amazons new cloud-powered browser Jon - - PowerPoint PPT Presentation

an overview of amazon silk
SMART_READER_LITE
LIVE PREVIEW

An Overview of Amazon Silk Amazons new cloud-powered browser Jon - - PowerPoint PPT Presentation

An Overview of Amazon Silk Amazons new cloud-powered browser Jon Jenkins Twitter - @jonjenk jjenkin@amazon.com OReilly Velocity EU November 8, 2011 Web Page Complexity is Steadily Increasing Amazons Home Page ~1996 2 elements 1


slide-1
SLIDE 1

An Overview of Amazon Silk

Jon Jenkins Twitter - @jonjenk jjenkin@amazon.com O’Reilly Velocity EU November 8, 2011

Amazon’s new cloud-powered browser

slide-2
SLIDE 2

Web Page Complexity is Steadily Increasing

slide-3
SLIDE 3

Amazon’s Home Page

~1996

2 elements 1 domain 10 kilobytes

slide-4
SLIDE 4

Amazon’s Home Page

~1996 2011

2 elements 1 domain 10 kilobytes 96 elements 12 domains 530 kilobytes

slide-5
SLIDE 5

Amazon Home Page Size Growth

Total Elements Total Kilobytes

20 40 60 80 100 120 1996 1998 2000 2002 2004 2006 2008 2010 100 200 300 400 500 600 700 1996 1998 2000 2002 2004 2006 2008 2010

slide-6
SLIDE 6

HTTP Archive Data

Total Elements Per Page Total Kilobytes Per Page

68 70 72 74 76 78 80 82 84 86 Nov-10 Oct-11 600 650 700 750 800 850 Nov-10 Oct-11

slide-7
SLIDE 7

HTTP Archive Data

HTML Elements Per Page HTML Kilobytes Per Page

1 2 3 4 5 6 7 8 Nov-10 Oct-11 32 33 34 35 36 37 38 Nov-10 Oct-11

slide-8
SLIDE 8

HTTP Archive Data

Javascript Elements Per Page Javascript Kilobytes Per Page

2 4 6 8 10 12 14 16 Nov-10 Oct-11 60 80 100 120 140 160 Nov-10 Oct-11

slide-9
SLIDE 9

HTTP Archive Data

CSS Elements Per Page CSS Kilobytes Per Page

1 2 3 4 5 Nov-10 Oct-11 22 23 24 25 26 27 28 29 30 31 Nov-10 Oct-11

slide-10
SLIDE 10

HTTP Archive Data

Images Per Page Image Kilobytes Per Page

45 46 47 48 49 50 51 52 53 54 Nov-10 Oct-11 400 420 440 460 480 500 520 Nov-10 Oct-11

slide-11
SLIDE 11

HTTP Archive Data

Number of Domains Per Page

2 4 6 8 10 12 14 Nov-10 Oct-11

slide-12
SLIDE 12

HTTP Archive Data

Average Page Speed Score

79 80 81 82 83 84 85 86 Nov-10 Oct-11

slide-13
SLIDE 13

Are we losing the battle? war?

slide-14
SLIDE 14

Migration to Mobile Makes the Problem Even Worse

slide-15
SLIDE 15

Desktop Laptop Tablet Phone CPU Quad 2.1Ghz Dual 2.1Ghz Dual 1.2Ghz Dual 1Ghz Memory 4GB 3GB 1GB 512MB Disk/Flash 1000GB 320GB 32GB 8GB Network 1Gbps 54Mbps 54Mpbs 10Mbps

What do you get for $500?

slide-16
SLIDE 16

The Network Matters Power Saving Mode (PSM) degrades wi-fi performance Radio Resource Control (RRC) degrades cellular performance

slide-17
SLIDE 17

Mobile Platform Load Times

Amazon Detail Page Load Times By Platform (milliseconds)

2000 4000 6000 8000 10000 12000 Sunday Wenesday Saturday Desktop Smart Phone iPad

slide-18
SLIDE 18

Complex Web Mobile + = L

slide-19
SLIDE 19

Complex Web Amazon Silk + = J

slide-20
SLIDE 20

What Makes Amazon Silk Different?

slide-21
SLIDE 21

The Architecture of the Web

slide-22
SLIDE 22

The Architecture of the Web

slide-23
SLIDE 23

Introducing Silk

slide-24
SLIDE 24

A Tale of Two Web Pages

www.google.com 15 elements 238 kilobytes 2 domains 1.951 sec load time www.cnn.com 175 elements 1,737 kilobytes 23 domains 11.181 sec load time

slide-25
SLIDE 25

Dynamic Split

slide-26
SLIDE 26

Dynamic Split

Kindle Fire EC2

slide-27
SLIDE 27

Dynamic Split

Kindle Fire EC2

slide-28
SLIDE 28

Dynamic Split

Kindle Fire EC2

slide-29
SLIDE 29

Amazon Silk Technical Details

slide-30
SLIDE 30

AWS

Optimizers Optimizers Browser Frontend ser WebKit

SPDY

Internet

HTTP(S)

Optimizers Browser Backend Distributed Cache

Silk Architecture

slide-31
SLIDE 31

Browser Frontend ser WebKit

Client Architecture

Rendering engine is Webkit JavaScript engine is V8 The application layer is built from scratch Automatic user agent mode Flash 10.3 support Falls back to local rendering on failure

slide-32
SLIDE 32

Server Architecture

Infrastructure is hosted on AWS No user-identifiable data logged Logs not kept for more than 30 days Optimizers will evolve over time

AWS

Optimizers Browser Backend Distributed Cache

slide-33
SLIDE 33

Optimizers

Image optimization Content minification Content compression Content push via page indexes Predictive rendering

Optimizers Optimizers Optimizers

slide-34
SLIDE 34

Protocols

Silk client and server communicate via SPDY

Multiplexed concurrent streams Request prioritization Compression Encryption Server push

Silk server communicates via HTTP(S) Silk client uses HTTP(s) when turned off SSL is always routed direct to origin

SPDY HTTP(S)

slide-35
SLIDE 35

Info for Web Developers

User Agent Strings At Launch

  • Mobile

Mozilla/5.0 (Linux; U; Android 2.3.4; en-us; Cloud9/1.0.0) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1

  • Desktop

Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_6_3; en-us; Cloud9/1.0.0) AppleWebKit/533.16 (KHTML, like Gecko) Version/5.0 Safari/533.16

User Agent String After First Software Update

  • Mobile – Acceleration On

Mozilla/5.0 (Linux; U; Android 2.3.4; en-us; Silk/1.1.0-72) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1 Silk- Accelerated=true

  • Desktop – Acceleration On

Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_6_3; en-us; Silk/1.1.0-72) AppleWebKit/533.16 (KHTML, like Gecko) Version/5.0 Safari/533.16 Silk- Accelerated=true

  • Mobile – Acceleration Off

Mozilla/5.0 (Linux; U; Android 2.3.4; en-us; Silk/1.1.0-72) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1 Silk- Accelerated=false

  • Desktop – Acceleration Off

Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_6_3; en-us; Silk/1.1.0-72) AppleWebKit/533.16 (KHTML, like Gecko) Version/5.0 Safari/533.16 Silk- Accelerated=false

slide-36
SLIDE 36

Info for Web Developers

Original IP address propagated in x-forwarded-for header We strictly obey caching headers Caching reduces sites bandwidth costs for site owners Content is optimized, but never replaced

slide-37
SLIDE 37

Info for Users

Everything is encrypted when in acceleration mode Caching on Silk server saves storage on the client Content compression saves bandwidth Acceleration can be turned off in settings menu Flash can be turned off in settings menu Browser works in offline mode for local content

slide-38
SLIDE 38

What AWS Services Does Silk Use?

Elastic Compute Cloud Elastic Block Store Elastic Load Balancing Auto Scaling CloudWatch CloudFormation Identity & Access Mgmt Route 53 Simple Queue Service Simple Storage Service Elasticache

slide-39
SLIDE 39

Final Thoughts

slide-40
SLIDE 40

References

Wi-fi Power Saving Mode (slide 16) http://www.networkworld.com/research/2008/051208-wireless-power-standards.html Radio Resource Control (slide 16) http://stevesouders.com/ms/ Data about growth of mobile (slide 14) http://www.onlinemarketing-trends.com/2011/06/state-of-tablet-adoption-infographic.html http://www.kpcb.com/internettrends2011 Web Page Test Data for CNN (slide 24) http://www.webpagetest.org/result/111107_5T_23VNE/ Web Page Test Data for Google (slide 24) http://www.webpagetest.org/result/111108_JP_23ZW2/ HTTP Archive Data (slide 6) http://httparchive.org/trends.php?s=intersection Internet Wayback Machine (slide 5) http://wayback.archive.org/web/*/http://www.amazon.com