Re: The future of ICP_OP_HIT_OBJ

On Tue, 18 Nov 1997, Alex Rousskov wrote:
> Has anybody measured the actual improvement from using the ICP HIT_OBJ
> feature? That is, how often it happens, and how much response time is
> saved?

I would be really happy to see that! ;-).

Actually, at Hungarnet we use ICP_OP_HIT_OBJ. We use it because of
the lack of persistent TCP connections in squid-1.1.x.

I think it is useful if you have only one peer - either a parent or a
sibling since - in this case your cache won't be able to receive the same
object more than once.

If persistent TCP connections were fully supported in squid, then - I
think - ICP should should act as an object locator service and we should
drop supporting ICP_OP_HIT_OBJ.

In case of a leaf proxy that has only one parent, I see that currently 13%
of acknowledged pings are ICP_HIT_OBJ, 28% are ICP_HIT. Anyone else has
other experiences?

Cheers,
Bertold
Received on Tue Nov 18 1997 - 23:47:21 MST