Fred Destin

FOWA: in the early afternoon, an inside look at Google API’s was provided by a senior product manager who ran through some of the best API usage examples currently out there, including NASA weather mapping, 3D rendering of UK landmark buildings such as the Globe Theater, live fight plan GoogleEarth integration.  At code.google.com all the API’s are available for the technically minded amongst you.  I never got Sphere to work but had no trouble creating personalised search for my blog, so they must be doing something right if non tech users manage to use their APIs too.

An iteresting piece of insight was provided by the fact that the guys at housingmaps.com did the first google mashup by hacking maps and pushed Google to release the API’s to the community as fast as they possibly could.

He was followed by Werner Vogels, CTO at Amazon, to talk about S3 and EC2 and other Amazon services.  Rather than talk tech, Werner focused on competing on ideas.  The concept is of course that Amazon takes the complexity away from you to allow the entrepreneur to focus on their business idea and not the backend.  His posterchild is Doug Kaye of the Gigavox podcast application: just drop in your MP3 and publish the podcast with appropriate encoding and ad insertion. 

The key message is that Gigavox built very little itself but rather use the EC2 computing cloud, the Queue service and S3 storage.  Pay for what you need.  So focus on the business idea, build it cheap, grow easily.

The key argument is around business continuity and focus: how can you, as a startup, scale 100x, provide 4 nines uptime, survive a complete datacenter failure etc.  Hard disks crash by 6-8% every year.  The argument is this is making life too hard for business that need to iterate constantly.  Vendor lock-in however was not addressed in the talk, which I certainly feel is a major issue. 

Examples put forward include:

• Smugmug (125K users, 120M photos).  Saved $470K in first 7 monyhs of operation, expecting to save around $2M this year
• SecondLife, using S3 as a CDN (streaming 70GB a day) to achieve offload
• Enrackados.com (a JS OS operating inside the browser) that uses S3 as the backend and takes an environment snapshot when leaving the session, saved on S3 and restored on the next login

The second service is EC2 i.e. the elastic compute cloud used for load-testing, time of traffic-based scaling, simulation and analysis, rendering, SaaS platform.  People run traditional grid based apps on this (see RenderRocket) or archiving services (Hanzo:web).

Another example was natural-language search company PowerSet (described as the most hyped company on the planet).  Ratrher than build a massive webcrawl engine they use EC2.

The full array of services include queue services, storage (S3), compute cloud (EC2), search (Alexa), commerce and Mechanical Turk (where humans come back into the picture).

The same concept was promoted by the excellent Simon Warldey who is building a similar service inside Canon / Fotango called Zimki.

technorati tags:fowalondon2007, wernervogels, amazon, google, api, s3, ec2, zimki