Cloud Computing today• Resources
– compute power– storage– networking
• Infinite capacity
• All you can use on demand
• Only one requirement– credit card starting with $50
Previous limitations
1. Infrastructure costs2. Development challenges
Infrastructure costs today• Higher by X times 2, 5, 10 years ago
Buy Rent Cloud RentAmazon EC2
1 rack = 20 servers
$65K total cost
$5K a month + contract lock in
$1.5K per month
Connectivity $9K per year
included nil
System AdminRequired
Yes Yes nil
Notes On demand pay per hour $0.1/h
Development challenges on a Cloud
Scalability and Quality
1. Communication stack2. Persistence / Transactional safety3. Application framework4. Development tools, environment and process
Development challenges on a Cloud
• Then vs. now: dramatic reduction in• Development cost• Time-to-market
– 2-3 developers can deliver complex pilot in 6 months
• Why?• Open source is production quality:
Axis/CXF, Hibernate, Spring, JUnit / TestNG, Eclipse, Maven, SVN, Amazon AWS, MySQL, Linux, etc
• Agile Development process - no RUP, no ITIL
Industry trends – Rise of Agile & Open Source
Proof of concept in a small team:Product, Technology, Costs, Time-to-Market
Goal1) Build 2 (later more) vertical business
directories in 15 languages2) By automatically summarizing business
websites3) Using a highly configurable crawler and
analysis engine running on a cloud
Rent/Single Server vs. Compute Cloud
Alternatives 1 Instance 40 Instances on Compute Cloud
Inventories 100K 100K
Mins to execute 1 inventory
2 min 2 min
Days 139 days 3.5 days
Cost $350No long term fixed & ops costs
Solution - Product
1. Best images2. Title3. Best sentences: Who, What, Where4. Feature tags5. Geo Location6. Other meta data – email, phone, etc
Solution - Product
Solution – Technology (Scope)
• 65K businesses/websites• Websites crawled 6 levels deep• 50gb highly structured data set• Transactions / records
– All together ~1M records• 250K sentence tags• 150K image tags• 500K feature tag
– 800MB DB filled
Solution – Technology (Cloud)• Servers run – 40 (tested on 80)• Total days - 8• Compute Hours ~1400 – 59 days on 1 server
Inventories crawled & analyzed & tagged per hour
0500
1000150020002500
12_2
5_7
12_2
5_2
12_2
0_14
12_2
0_9
12_2
0_4
12_1
9_23
12_1
9_18
12_1
9_13
12_1
9_3
12_1
8_22
12_1
8_17
12_1
8_12
12_1
8_7
12_1
8_2
12_1
7_20
12_1
7_15
12_1
7_10
12_1
7_5
12_1
7_0
Inventories tagged
Solution – Costs
Demo
Conclusion: Now vs. ThenMajor shift
1. Unlimited cloud resources for small teams
2. Infrastructure cost – miniscule3. Reduced Time-to-market4. Lower Development cost
Product vs. Revenue models
• Consumer product – Advertising– Transaction fees– Subscriptions– Premium services
• Enterprise / Business product