> 1) Start downloading a big file (2MB) from a nearby server -> OK, as
> expected the file is transfered at 6K/second.
>
> 2) While this transfer is active, try to open a few small web pages, also
> on nearby server.
>
> Hoped-for reaction: the available bandwidth of 6k is shared more or less
> evenly between the outstanding requests.
>
> Observed reaction: the big download keeps running at 6k/sec and no other
> requests get serviced at all until the big one finishes.
The other requests are not serviced at all, or serviced extremely slowly?
The second is what already happens over a modem, and I would have expected what
squid would do. There is no priority given to earlier connections or anything
like that.
> Another question: how does delay-class on proxy A impact clients connected
> to proxy B, if A is the parent cache of B? Are prox B's ICP requests reated
> just as if it was a normal client?
ICP is not delayed, only TCP connections. Already cached objects aren't
delayed either.
David.
Received on Thu Nov 12 1998 - 18:47:37 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:02 MST