The Modular Approach to a Scalable Cloud
Grazed from ITBusinessEdge. Author: Arthur Cole.
Following up on my previous post regarding hyperscale infrastructure, I feel I should point out that once the decision to go hyperscale has been made, it will most likely take place in a Greenfield hardware environment. Unless you are already working with a state-of-the-art data facility, any attempt to convert complex, multiformat legacy environments will almost certainly lead to a morass of integration issues.
The key benefit to hyperscale is that it is both large and flexible, allowing data executives to craft multiple disparate data architectures completely in software. This is why current hyperscale plants at Google and Facebook rely on bulk commodity hardware. But as I mentioned last fall, the average enterprise does not have the clout to purchase tens of thousands of stripped down servers and switches at a time, and besides, all those components still need to be deployed, provisioned and integrated into the cluster, which takes time, effort and of course, money…


