Need help about SQUID access.log (How can I get the duration of an ICP query?)

Hi,

I am doing a project related to cooperative cacheing.
Now my system includes three SQUID machines(FreeBSD
with SQUID installed) connected to a set of Polygraph
machines who could simulate clients/servers for the
cacheing system.

Now I need the actual duration time for every ICP
query between two SQUIDs. I examined the access.log
file and found there're some records describing it:

1049144619.647      0 10.12.0.2 UDP_MISS/000 86
ICP_QUERY http://10.12.130.8/w08
2761e4.1f940629:7fffff9f/t01/_00000101.jpeg - NONE/- -

the second field(0 in this case) is the duration for
the message. 10.12.0.2 is the sibling of current SQUID
cache(10.12.0.1). This message means the sibling_SQUID
10.12.0.2 just asked 10.12.0.1 that whether or not it
has the above object. And the "UDP_MISS" told the
sibling that 10.12.0.1 doesn't have the object in
cache, so it's a ICP_query and generate a UDP message
to tell the sibling a miss.

My question is why the duration for these IPC_querys
are always 0 milli-seconds? Even though it could be
fast, I don't believe it can be that fast....

Or did I misunderstand the value of duration field? If
so, could you please tell me how can I get the actualy
duration for a specific ICP query between two sibling
caches?

Thank you so much!

Ming

PS: I also tried to use the following commands of
dummynet to add some latency to the UDP messages, but
the records is still giving me 0 milli-second for
ICP_Query.

ipfw add pipe 1 UDP from any to any
ipfw pipe 1 config bw 500Kbit/s queue 10 delay 10ms

Where did I do wrong?



__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com