These are the old pages from the weblog as they were published at Cornell. Visit www.allthingsdistributed.com for up-to-date entries.

March 19, 2004

Feed Analysis: Requests vs. Bytes - Part I

The graphs I gave on March 5 about the visitors versus feed transfers was incomplete. At that moment I had discarded the outliers in the distribution (those who visited more than 1000 times in the two month period), but by now I know that was wrong. Below you will see that numbers revisited combined with a view of actual bytes transferred.

In the first graph we examine the number of visits the individual users made during the two months visit. Here we look at the cumulative distributions for the number of unique users, the number of requests made as a percentage of the overall requests and the total number of bytes transferred in those requests. There is a relation of course between the feed requests & data transferred and the chance of a NoChange response as shown in the previous posting. A Success response will result in transfer of 45 KBytes while a NoChange is 250 bytes in size. Given that the lower request rates have a higher success percentage we see that the line for the bytes transfer rises faster than the #requests, but smoothens out quickly, until it hits a few spikes at the end of the graph.

If we categorize the data we see that following for number of requests and data transferred:

            

Posted by Werner Vogels at March 19, 2004 11:56 AM
TrackBacks

Comments