Those who design and operate content delivery networks look for speed, dependability, and consistency. Though several methods are required to achieve these objectives, this paper focuses on those that the authors find “technically interesting.”
To be more precise, the paper “walks through the steps” required to be completed from the moment a request for some content is made, through some browser, to its actual delivery. Being an empirical researcher in algorithm analysis myself, I fully agree with the authors that “algorithm design does not end when the last theorem is proved.” In fact, theoretical results are often contradicted experimentally; for example, an asymptotic bound may not be realistic over a finite range over which a computer experiment can be conducted. The authors are quite right in assessing that speedy, scalable, and cost-effective implementations require practical considerations. They have accordingly focused on the translation of algorithms that are of recent demand in the industry.
Some of the algorithms discussed in the paper, along with their empirical benefits, are global load balancing, the stable marriage problem, consistent hashing, bloom filters, and overlay routing.
The paper contains useful material and will certainly draw interest among researchers working in the client/server environment.