Re: public_interest field

From: Alex Rousskov (
Date: Thu Oct 11 2001 - 13:28:31 MDT

On Thu, 11 Oct 2001, Ahuja Sadhna (NRC/Boston) wrote:

> I am interested in testing peering capabilities between two
> caches. So, I would like to find out how can polyclt processes running
> on two different machines make some common requests to their respective
> caches. I have looked at the "proxy cache cooperation" thread
> ( ), but
> I am not sure I understand it well.
> The Polygraph docs
> ( ) say:
> "...each robot has to maintain two ``worlds'' of URLs. The first,
> ``public'', world is shared among all the robots of all polyclt
> processes. The second, ``private'', world is not shared. For every
> request, Polygraph decides which world should be hit based on robot's
> public_interest configuration field. "
> Does it mean that the "public" world is shared by robots of all polyclt
> processes running on the same machine?

Public world is shared among all the robots of all polyclt processes
on all machines in all buildings on all planets participating in a

Addressing schemes aside, Polygraph does not know or care about
"machines". Virtually all algorithms work with individual agents such
as robots or servers. Processes, machines, etc. are just "packaging".

> Or is it a fixed set of URLs that is shared across multiple machines?

The set is shared, but it is not necessarily fixed. The size of the
public world (public working set size) is determined by several
factors, including working_set_cap() function call in PGL.

> Is the value of public_interest a probability relative to the recurrence
> ratio?

Public_interest is the probability that public world is selected when
the choice between public and private world is made. Public_interest
is not related to the recurrence ratio. The effects of the two
probabilities are independent (you can have hits in both worlds) as
long as both worlds have at least one object (so that a hit can be



This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:20 MST