Cloudonomics: The Business Value of Cloud Computing, + Website

CHAPTER 20

Dispersion Dilemma

Let’s recap the logic of the last few chapters. We’ve argued that milliseconds and microseconds can mean millions of dollars. Whether trading equities or serving coffee shop customers, time is often of the essence. To reduce response time, we pointed out that you could leverage parallelism, where many hands make light work, and/or geodispersion, a sort of zone defense spread out over the planet.

In the case of response time reduction through parallelism, there are diminishing returns after early gains if you are attempting to achieve these gains yourself. However, the cloud can be a huge benefit, because, the pay-per-use model means that rather than suffering through a long elapsed time due to a small number of resources, a short elapsed time—leveraging many resources in parallel—can be achieved at no additional cost.

In the case of response time via dispersion, a roughly similar analysis holds. Increasing the number of delivery points can reduce time due to network latency by reducing the average or worst-case distance between a user and a service node. Again, using your own funds to pursue these gains rapidly becomes prohibitively expensive, because eking out smaller and smaller latency improvements requires rapidly growing investments. Again, the cloud can help by providing pay-per-use access to resources.

Generally, consolidation has benefits, such as aggregate demand smoothing, ability to leverage parallelism, and economies of scale, but dispersion ...