Chapter 16. Rapid Flexible Scalability

How often have you looked at the logs that are monitoring a system's usage and seen minimal use of memory, CPU, or other system resources? Servers are built to handle anticipated peak loads. Normal practice and the natural tendency to anticipate the worst and/or cover your behind usually mean that the highest peaks that a server can handle are almost never encountered. This is expensive, both in terms of capital costs for hardware and licenses and operational costs such as electricity, cooling, floor and rack space costs, support staff and other incidental costs. It also imposes significant burdens on the operational staff supporting the machines. The more machines one uses, the more likely it is that one of them will break. In keeping with Murphy's Law, this break will, of course, occur at 4:55 PM on the Friday before a holiday weekend.

The solutions associated with cloud computing have been around for some time but have typically been seen as immature...