Thanks for the helpful discussion on delay pools previously. This is
what I am implementing:
# acl for dialup users:
acl localdelayed 203.22.103.1-203.22.103.127/255.255.255.255
# apply class2 delay pools:
delay_class2_access allow localdelayed
# set the overall bucket size to 64kb (we have a 512k link so this is
the max)
delay_class2_aggregate_max 64000
# fill the bucket at 64kb per second
delay_class2_aggregate_restore 64000
# set a larger max bucket size (16kb) so that web pages etc should be
faster
delay_class2_individual_max 16000
# limit any downloads after this to 8kb per second
# - modem/64k ISDN users shouldn't notice any speed degradation as
they
# can't download faster than this (well, forgetting compression)
delay_class2_individual_restore 8000
Any comments?
A few things I'm unsure about:
1. To what extent does the 16kb bucket size actually make short
downloads faster? Won't the modem still be the bottle neck?
2. How can I remove the aggregate settings totally? I don't care about
the aggregate bucket, at least for the moment. squid.conf recommends
setting aggregate_restore to -1 but what about aggregate_max?
3. How is the acl defined here different to a subnet definition, say
203.22.103.0/255.255.255.128?
TIO,
Chuck
Received on Thu Nov 19 1998 - 04:46:21 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:10 MST