How To Create Markets With Technology

• Utility 2.0 (using cloud computing) is a huge departure from Utility 1.0 where you provision your own environment
• Virtualization is a big enabler of Cloud Computing
• Look for a big land-locked demand for a service and unleash it with Cloud Computing
• SaaS: Only an account sign up is required to use the service. Examples:Yahoo! Mail, ebay, flickr.
• PaaS: Platform as a Service: Making implementation for apps targeting similar platforms MUCH simpler. It targets niche markets making them very efficient and easy to deploy. The problem is that if you have existing code (C, C++) wil not work in PaaS. You might need to go IaaS. Examples of PaaS: AppEngine, Force.com
• IaaS: Infrastructure as a Service: You have your instance of OS that meets your requirements. No network latency. AWS, GoGrid, Mosso.
• Hadoop: Game changing for number crunchers and large web apps
• Cloud computing is a technology and operational leap
• IaaS: Mostly used by IT people
• PaaS: Mostly used by developers
• SaaS: Mostly by business users

• Use case: Building a Web App. Not serious about infrastructure and architecture yet at the beginning. At this stage the cloud is not for you 
• If you only need one server, you shouldn’t be using the cloud (One blog can cost you a 700% idiot tax in the cloud)
• Cloud computing: Bridging the gap between super expensive service companies that set up data centers and the sole server setup. Addresses mid-tier market where people want to build scalable applications without the upfront cost of the past
• Michael from Microsoft: Web hosting model is a commodity competing in price while cloud computing is not. That’s the reason why cloud computing services are able to charge you the 700% idiot tax. 
• Application use case: When the application gets enough traffic, then the needs are different and the cost of downtime or slow load times is bigger. 
• Dynamic Data in Social networks in the cloud. All this information needs to be available instantly which creates a challenging scalability issue. How do Social Networks address the issues? 
  • MemCache for serving data. Makes it less costly to serve page. 
  • Two primary modes to storing data in the cloud. 
    1. Consistent (Primary storage of data is one single primary store and DB always knows the latest information. Transactional). Fighting for access to a value that might be changing faster than it can be committed.
    2. Eventually consistent. Multiple DB that store information. One stores updates but never used for reads and it pushes copies to read-only databases. Not ideal but cost efficient
  • Number crunching: The most efficient way to crunch is not to do it ins  single database instance but to take your code and push it into the data and let it crunch the data. Hadoop manages most of this process. At the core, Hadoop maps the set of data into small pieces and reduces the computing time. Once al the pieces are computed then it maps it back to the original state.
  • The issue is how to get the data into the Cloud.
  • Google is traveling with the data across the country to transfer it because it is cheaper.
  • Real issue is not in the computing power but in the hard-drive performance

• Achieving 5 nines is difficult (requires geographic separation)
• At this point you can count in 3-nines for SLA
• Most products are still on beta
• 5-nines = downtime of 5 minutes a year
• Even though a cloud vendor might not be SLA of 5-nines, you can code your app to creat 5-nines for yourself
• Games: Copy files to go-grid. Store pointers to those files in the DB and if one fails then you flip a switch and start serving files from the other cloud vendor.
• Access to HW very quickly. 
• Automated procedure to replication on top of “disposable” HW
• Benefits of redundancy: Scale instantly across the world. If one goes down then you can redirect traffic to other locations

How do you manage your procedure without face-to-face meetings?