In his recent article Pod-Scale vs Warehouse-Scale Computing, Phil Wainewright talks about the differences between how Oracle and Microsoft view cloud and SaaS computing infrastructure and how Google views it. Oracle talks about tailoring instances of their on-demand applications to different customers, time zones and functions. This sounds like the old ASP model to me. Google views their services as components of a global operating system. Their data centers are their versions of my desktop or server. The operating system uses a single file system, database and number crunching system. The difference is that Google uses thousands of machines as their virtual computer where I use one. The services have been distilled into a single homogenized set of primitives that are distributed. When Google updates one component the whole system benefits, hence all customers benefit. This is one of the great advantages of SaaS and using a multi-tenant, single instance approach. Oracle and Microsoft still think about separate systems tuned for separate applications. This looks like a way to maintain their legacy.
So who is right? It’s tough to argue with the success of Oracle and Microsoft to date. Whether they are just packaging their old applications as SaaS or they are creating new ones to take advantage of new architectures is debatable. As with all things in SaaS, everyone has a different approach. I think that the Google approach is more scalable and easier to update. This is real cloud computing, in my opinion, and the great advantage of SaaS. With things moving at web speed, I think this is the better approach.
photo credit Marcin Wichary