Measuring and improving origin offload and resource utilization in caching systems

A method for calculating offload associated with cacheable objects delivered by one or more cache servers for a content provider, the method comprising:

• with the one or more cache servers in a content delivery network, receiving client requests for the cacheable objects associated with one or more hostnames associated with a content provider;
• generating, with respect to a configured time period:
• (i) one or more statistics indicating how often the one or more cache servers had to contact another server due to cache misses for a plurality of the cacheable objects requested by clients during the configured time period, and (ii) an estimate of the number of unique cacheable objects amongst the plurality of the cacheable objects requested by clients that resulted in contacting the another server due to cache misses in (i);
• based at least on the one or more statistics from (i), calculating an observed origin offload for the configured time period;
• based at least on the estimate from (ii), calculating a maximum possible offload for the configured time period;
• and,
• providing the observed origin offload and the maximum possible offload for comparison in any of:
• user interface and an analytical tool;
• wherein the estimate in (ii) is determined by a set of steps comprising:
• providing a first function that maps cache keys associated with the plurality of cacheable objects to points in a set of uniformly distributed points in a range associated with the output of the first function;
• during the configured time period, upon observing a forward request for a particular cacheable object of the plurality of cacheable objects, obtaining the corresponding mapped point for the particular cacheable object, where the forward request results from a cache miss for the particular cacheable object that requires contacting another server for the particular cacheable object;
• maintaining a data structure storing the K-th sized mapped point associated with the forward requests during the configured time period, K being a configured value;
• and,
• generating the estimate of the number of unique cacheable objects associated with the forward requests during the configured time period, at least by applying a second function to the K-th sized mapped point.