[squid-users] accelerator farm: optimizing the sibbling_hit

Hi,
I am busy optimizing an accelerator farm using 2.4.4.
It currently contains the following *easy* patches:
- remove updates into client_db, this makes sure that when you
  have a major site, your accelerator won't run out of memory.
  This means commenting out the few calls to clientdbUpdate
- Allow refreshes to be done by peers.
  This means removing a test in neighbors.c, function
  peerAllowedToUse, test: request->flags.refresh

I am now thinking of the next ICP improvement, because having the
same aging criteria on all accelerators will give the following
problems: it will expire on all servers at exactly the same time.
When you have a high profile website, you can expect all
individual accelerators to connect to the same site at the same
time. Nice when there are some generated pages...
So I am thinking to make each accelerator expire at another time.
F.I.:
1 page has an expire at 13:00
accelerator 0 has a patch that expires it 0 seconds before the
real expire time
accelerator 1 has a patch that expires it 3 seconds before the
real expire time
accelerator 2 has a patch that expires it 6 seconds before the
real expire time

In a normal situation accelerator 2 will need to refresh the page
6 seconds before the real expire, and will therefore be the first
to request the page. Accelerator 1 will find out 3 seconds later
that accelerator 2 already has the page, and will use that new
expire time. Accelerator 0 can feed either from accelerator 1 or
2.
(Yes, it also means I have changes icp_refresh timings)

My question: is anybody else already doing this same thing?
(That is: creating an accelerator farm with optimized
sibbling_hits?)

-- 
program signature;
begin  { telegraaf.com
} writeln("<ard@telegraafnet.nl> SMA-IS | Geeks don't get viruses");
end
.