I want to get one thing out of the way right upfront. It is possible to scale the LAMP stack to hundreds of millions of page views per month and beyond. But that largely involves unnatural acts. By that I mean developing lots of custom backend code that has no visible value to an enduser of the site or service. My third principle of cloud computing is that code scales under normal circumstances without unnatural acts. In the ideal cloud scenario you can take a Ruby on Rails or Django app, drop it into the cloud and not have to worry. Obviously one needs to put some constraints - the ‘normal circumstances’ - on that. At least for the next few years, I think you should expect to be able to scale up to 100 million page views per month without contortions and to handle traffic spikes up to 10x average traffic levels. For sites that do less than 1 million page views per day even higher spikes should be handled, ie if you do one thousand views a day and the suddenly get to the front of digg you should still be OK. All of that should happen without requiring any intervention by the developer, eg no requirement to log in and add capacity. I believe that we have the tools to make this work for a large number of sites and services. This will be hugely liberating for developers as long as it is done without lock-in.

I want to get one thing out of the way right upfront. It is possible to scale the LAMP stack to hundreds of millions of page views per month and beyond. But that largely involves unnatural acts. By that I mean developing lots of custom backend code that has no visible value to an enduser of the site or service. My third principle of cloud computing is that code scales under normal circumstances without unnatural acts. In the ideal cloud scenario you can take a Ruby on Rails or Django app, drop it into the cloud and not have to worry. Obviously one needs to put some constraints - the ‘normal circumstances’ - on that. At least for the next few years, I think you should expect to be able to scale up to 100 million page views per month without contortions and to handle traffic spikes up to 10x average traffic levels. For sites that do less than 1 million page views per day even higher spikes should be handled, ie if you do one thousand views a day and the suddenly get to the front of digg you should still be OK. All of that should happen without requiring any intervention by the developer, eg no requirement to log in and add capacity. I believe that we have the tools to make this work for a large number of sites and services. This will be hugely liberating for developers as long as it is done without lock-in.