On Mon, Dec 28, 2009 at 2:04 AM, Shawn Walker <[email protected]> wrote:
> image-update from 129 -> 130 using gate tip bits:
>
> URL                   Good   Err  Conn Speed    Size       Used  CSpeed
>  Qual
> pkg-na-2.opensolaris.org 23264  0    5    518 kB/s 416.25 MB  True 0.088275
> 50

Here's what I see...

URL                   Good   Err  Conn Speed    Size       Used  CSpeed  Qual
pkg-na-2.opensolaris.org 117    0    2    43 kB/s  696.68 kB  True
1.863239 -29217
pkg.sun.com           161    0    16   80 kB/s  62.20 MB   True  0.087710 32
pkg.opensolaris.org   24688  0    22   60 kB/s  436.34 MB  True  0.088523 63
192.168.1.102         0      23   0    0 B/s    0.00 B     True  0.000000 -6571

Prior to starting this I created flows corresponding to
pkg.opensolaris.org and pkg-na-2.

# flowadm add-flow -l ath0 -a remote_ip=192.9.76.17 na2
# flowadm add-flow -l ath0 -a remote_ip=72.5.123.21 pkg

Then I captured the output with something similar to

# flowadm show-flow -i 1 -s pkg | tee /var/tmp/pkg
# flowadm show-flow -i 1 -s na2 | tee /var/tmp/na-2

Then a bit of post-processing shows more informative throughput data.
Note that there are large amounts of time when no network activity is
happening and I exclude those.  The columns below are number of KB/s
(0 -> 0 - 99; 100 -> 100 -> 199, ...), number of occurrences, and a
histogram of number of occurrences

# nawk '$1 != "FLOW" && $3 != 0 {printf "%d\n", $3 / 100000}' \
     /var/tmp/pkg | sort -n | uniq -c \
    | perl -nle 'split(/\s+/);
        printf "%4d %3d %s\n", $_[2] *100, $_[1], ("*" x ($_[1] / 10))'
   0 540 ******************************************************
 100 289 ****************************
 200 204 ********************
 300 198 *******************
 400 169 ****************
 500 145 **************
 600 115 ***********
 700  79 *******
 800  29 **
 900  18 *
1000  11 *
1100   4
1200   1
1300   1

The same for pkg-na-2:

   0   8
 100   1
 400   1

Note that in the past I have noticed my bandwidth being throttled for
large (e.g. .iso image) downloads.  I have not analyzed the collected
data to see if that happened here or not.

-- 
Mike Gerdts
http://mgerdts.blogspot.com/
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to