> By Frédéric Aidouni
Using the data presented in the paper Ten weeks in the life of an eDonkey server we computed the number of queries of each type managed by the server each hour. For each hour, we then computed the ratio between the number of answers to source-search query divided by the number of answers to keyword-driven queries. In other words, we observe the average number of answers to source search queries generated for each answer to a keyword-driven query. We then observed this number for the first hour of all days, second hour of all days, etc, thus obtaining the plot above.
The inset plot was obtained similarily, but by directly considering the number of answers given to keyword-driven queries.
Each blue dot represents an hour of measurement, and the red dots give the average for each hour of day.
In both plots, a clear day/night phenomenom appears, indicating that the large majority of client are in similar timezones (the ones of western Europe, probably).
However, the plots have very different shapes, which we may interpret as a consequence of user behavior in reaction to server load (which make it slower).