On Tue, 2 Apr 2002, Peter Backx wrote:
> I have some more questions on customizing the content that is
> requested by clients (robots) and that a server can supply. I've
> checked the PGL manual and I can't seem to find what I'm looking
> for:
>
> * Is it possible to limit the total number of different objects a
> server offers? For instance I would like to limit the entire
> "data-set" that gets requested to very few objects (like 50)
>
> * And on the client side: is it possible to limit the different
> objects that a client requests. This would be of course limited by
> the above mentioned server limit, but I would like to narrow it
> down a little more (if possible)
Yes, it is possible. You need to fiddle with three simulation
concepts: the size of the working set, the sliding speed of the
working set, and a hot subset of the working set.
For working set size (WSS) discussion, see working_set_length() and
working_set_cap() PGL calls:
http://www.web-polygraph.org/docs/reference/pgl/calls.html#call:docs/reference/pgl/calls/working_set_length
The speed at which working set slides is affected by the rate at
which robots are requesting new objects. That rate is determined by
recurrence ratio of a robot multiplied by the robot request rate:
http://www.web-polygraph.org/docs/reference/pgl/types.html#type:docs/reference/pgl/types/Robot
Hot subsets in Object Popularity Model (PopModel PGL type) can be used
to instruct robots to mostly request a subset of a working set:
http://www.web-polygraph.org/docs/reference/pgl/types.html#type:docs/reference/pgl/types/PopModel
Here is a draft plan to get a desired constant working set effect
(i.e., a constant size working set with zero sliding speed) and
"narrow it down a little more" with popularity model settings
1. limit the working set size using working_set_cap(50) PGL call
2. set robot recurrence to 100%; set public interest to 100%
or at least 90%; set robot's popularity model (pop_model)
so that hot_set_prob is high and hot_set_frac is as low as
you want it to be
3. in a "fill" phase, grow the working set size until it
reaches the limit and gets frozen;
see how PolyMix-4 does it in the fill phase;
use 0% recurrence factor for this phase to speed things up
4. in a "measurement" phase, set recurrence factor to
100% so that virtually no new objects are introduced into
the working set (now frozen at 50 objects level)
You may see some side-effects of a very small working set size and
very high recurrence ratio. For example, if your measured hit ratio is
also 100%, your robots will not be able to synchronize phases because
they will not be able to talk to the servers.
> Although these questions aren't very important in a "real life"
> case where you want to test a cache, I want to make these changes
> to the content in order to verify my practical impementation with
> a previously made theoretical study we did.This study was rather
> limited in available parameters (and thus possible content and
> requests), because of practical limitations when solving ILP
> problems.
Please share your results or problems in getting this to work. Other
people may need to model similar environments so your experiences may
be of interest to the general Polygraph public.
Alex.
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:22 MST