Alex --
Thanks for the response, I really appreciate it. Here
are a few follow up questions.
> For direct requests, a robot picks a random server
> from its "origins" array.
What sort of distribution does it use in this
selection process? Is it more likely to stay on the
server it last requested a "page" from or is any
server equally likely? or is this driven by the
pop_model? (see below)
> For embedded objects (e.g., inlined images),
> the robot extracts the URL from markup and follows
> it like a real browser would. In PolyMix, all
> embedded objects reside on the same server as their
> HTML container.
Am I right in thinking the following:
Once the robot loads the "page" it then waits for its
next request time based on its request rate and
typically GETs another page. Then, driven by the
pop_model, recurrence and embed_recur, it chooses a
new "page" to GET. This "page" is associated with a
server and if this server is not the same as the last
server then the Robot does a DNS request to get the IP
of the new server. If this "page" is associated with
the same server then the Robot does not do a new DNS
request.
i.e. does the popularity model effectively drive
server selection or is server selection independent of
the popularity model?
> Please let me know if your environment requires
> different connection/DNS use patterns.
I think what you have is probably fine, I'm just
trying to understand it.
Thanks!
Andrew
__________________________________________________
Do You Yahoo!?
LAUNCH - Your Yahoo! Music Experience
http://launch.yahoo.com
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:23 MST