Monday, December 11, 2006

Scaling up is not an option

In the Web 2.0 world there are two critical capabilities. Innovation in assembly which involves taking the best of breed components and functionality from across the web and integrating them in innovative ways into your own platform to create a unique and hard to recreate database of information. The other absolutely critical function is operational capability. The ability to scale your platform at hyper-accelrated rates is essential. When your platform reaches the inflection point and popularity and traffic skyrockets you have to be able to deploy the necessary compute and storage resources to meet customer demand. Deteriorating performance and long response times will drive away customers.

The ability to scale and deploy hardware resources has been coming sharply in to focus. Greg Papadopoulos, Sun's CTO, has garnered a lot of attention from a statement he made in his recent blog. He stated that the world only needs 5 computers, but then preceeded to name seven candidates. One of the companies missing from the list has a tremendous track record in scaling. It has done so successfully as the Internet itself has gone through hypergrowth. That company is AOL. The Papadopoulos list of hyper-scale processing grids includes:

  1. Google
  2. Microsoft Live
  3. Yahoo
  4. Amazon
  5. eBay
  7. The Great Computer of China
If AOL is going to rise in to this list of major players and avoid being assimilated by one of them it has to quickly learn to take the inate scalability capability that it has developed and externalize it. The best model in the list is Amazon. It has taken the infrastructure on which it runs its own business and monetized it. Their pay for use computing, storage and services model benefits Amazon by not only bringing in additional revenue but also enabling Amazon to benefit from additional economies of scale.

When you first look at Amazon's charging model the small 10 to 20 cent charges look reasonable. When you scale it up you realize that you may quickly end up spending a lot of money with Amazon to run your Web operations. What you have to remember is that you are avoiding significant expense by not having to build your own infrastructure to handle the peaks of demand. When you consider those contingencies the Amazon option becomes very attractive. Significantly for Amazon, this pricing model leverages the network effects of other Web innovators.

AOL has to be one of the contenders with the potential to rise in to this list. Their infrastructure is extensive. As an example AOL routinely blocks two to three billion spam emails every day. The multimedia infrastructure handles the global distribution of the Live8 concert at very short notice. The capability is there. The launch of, the creation of AIMPages and the rapid improvements to and, indicates that the potential is there. AOL needs to not only find ways to monetize the forward face of AOL but also the underlying production capacity that it has spent years investing in.

The divestiture of the access business and the new found focus on an advertising supported business model positions AOL to grow in the future. The proven ability to scale gives AOL the potential to grow in to the Papdopoulos list of compute grids. What are your thoughts? Who else is missing from the list? Join the conversation and leave a comment below.

No comments:

Post a Comment