Werner Vogels @ FOWA Feb 07

Loading...

Flash Player 9 (or above) is needed to view presentations.
We have detected that you do not have it on your computer. To install it, go here.

0 comments

Post a comment

    Post a comment
    Embed Video
    Edit your comment Cancel

    2 Favorites

    Werner Vogels @ FOWA Feb 07 - Presentation Transcript

    1. Web Scale Computing: Compete on Ideas, Not Resources Werner Vogels Fast Forward to 33 minutes CTO – Amazon.com 2 3 4 What if… What if… • Launching a new business on the web was simple? • Launching a new business on the web was simple? • You only had to focus on “the business”? • You only had to focus on “the business”? • You could manage growth more easily? • You could manage growth more easily? What if you only had to compete on ideas? 5 6 1
    2. PUSH VS PULL • Demand is anticipated • Demand is uncertain • Top down design • Emergent design • Centralized design • Decentralized • Procedural • Modular • Tightly coupled • Loosely coupled • Resource centric • People centric • Restricted participation • Open participation • Limited re-engineering • Rapid incremental innovation 7 8 Forces Driving Alternative Resource Models DESIGN • Increasing Uncertainty l • Growing Abundance de • Intensifying Competition o M REFINE DEPLOY • Growing Power of Customers h • Greater Focus on Learning and Improvisation us P EXECUTE & MONITOR 9 10 Resource Management in an Uncertain World FIND • Acquire resources on demand l • Release resources when no longer needed de • Pay for what you use o CONNECT REFLECT • Leverage other’s core competencies lM l Pu INNOVATE 11 12 2
    3. A New Web Business using the Push Model The 70 / 30 Switch Undifferentiated Successful Your Idea “Heavy Lifting” Product 30 % of time, energy, and dollars on differentiated value creation Hardware Costs Software Costs Maintenance Expertise Load Balancing Scaling and Physical Growth 70 % of time, energy, and dollars Costs to run idle servers on undifferentiated heavy lifting Bandwidth Management Server Hosting 13 14 How can you ever build an application that needs to • scale to 100% - 1000% increase popularity? • provide 4 nines uptime? • survive a complete datacenter failure? • survive a network partition • While keeping cost low at the same time? 15 16 The Reliability of Hard Disks Failure Trends in a Large Disk Drive Population Chapter 4: Priorities – Scale Later – It is too hard to get right Eduardo Pinheiro, Wolf-Dietrich Weber and Luiz André Barroso 17 18 3
    4. Challenges to just get the infrastructure right Resources in the Pull Model: Web-Scale Computing • Networks, hardware, operating systems, databases, Scalable Infrastructure that allow your applications middleware all fail all the time to meet infinite demand, cheaply and reliably • How to beat the CAP theorem • Turn huge fixed costs into variable • Datacenters also fail (tornados, heat waves) • Scale up and down as your business does • To use scale as a cost-effective advantage you need really large numbers • Pay as you go • Operations is just plain hard, large & small • Leverage other’s core competencies • Invest intellectual bandwidth to build the infrastructure • Focus on your Idea • Invest funds for something that may not happen (yet) 19 20 Amazon S3, EC2 & SQS Amazon Simple Storage Service (S3) Scalable – Increase or decrease capacity in minutes Storage for the internet via web service calls • Cost-Effective – Low rate, pay-as-you-go Private and public storage options • Reliable – Runs on Amazon’s proven infrastructure Simple – SOAP- and REST-based Web Service Calls $.15 per GB/ per Month to store data • Compatible – Use Amazon EC2 and S3 together to receive free data $.20 per GB/ per Month to transfer data • transmission between services, decreased latency, and consistency. Use Cases: Unlimited Data Storage, Media Sharing/ • Distribution, Archiving, Server Back-ups 21 22 Smugmug’s growth 140 120 millions of 100 photos 80 60 40 20 0 2002 2003 2004 2005 2006 23 24 4
    5. Smugmug & Amazon S3 • Saved $474,000 in first 7 months of operation • Pre-S3: – Buying $40K/month of new hardware – Without S3 this would now be $80K/month • Post-S3: – Selling excess hardware on eBay – Costs even lower than projected • Expecting savings of $1M to $2M in 2007 25 26 27 28 S3 Explorer Jungle Disk Amazon Elastic Compute Cloud (EC2) filicio.us Virtual Computing Environment Scalable Capacity – Pay as You Go $.10 per Server Hour $.20 per GB of data transfer Use cases: Load testing, Time or Traffic-based S3 Firefox Organizer MyOwnDB Scaling, Simulation and Analysis, Rendering, SaaS Platform 29 30 5
    6. 31 32 AWS Product Family Infrastructure: Web Search: Amazon Simple Queue Service Alexa Web Search Amazon Simple Storage Service Alexa Web Information Service Amazon Elastic Compute Cloud Alexa Top Sites Alexa Site Thumbnail E-Commerce: Workforce / Workflow: Amazon E-Commerce Service Amazon Historical Pricing Amazon Mechanical Turk 33 34 http:/ /aws.amazon.com 35 6

    + carsonsystemscarsonsystems, 3 years ago

    custom

    964 views, 2 favs, 0 embeds more stats

    Werner Vogels of Amazon speaking at Future of Web A more

    More info about this document

    © All Rights Reserved

    Go to text version

    • Total Views 964
      • 964 on SlideShare
      • 0 from embeds
    • Comments 0
    • Favorites 2
    • Downloads 52
    Most viewed embeds

    more

    All embeds

    less

    Flagged as inappropriate Flag as inappropriate
    Flag as inappropriate

    Select your reason for flagging this presentation as inappropriate. If needed, use the feedback form to let us know more details.

    Cancel
    File a copyright complaint
    Having problems? Go to our helpdesk?

    Categories