From: Connary, Julie Ann (jconnary@xxxxxxxxx)
Date: Thu Dec 28 2000 - 13:05:24 GMT-3
Hi Curtis,
Is this not a custom queuing problem not a frame-relay traffic shaping problem?
since for 128K - that is 1/12 of the bandwidth of the 1.544Mb total or
about 8% of the bandwidth. the
default queue byte size is 1500. So if you want queue 1 to service http and
only have 8% of the bandwidth and
the rest to have 92% of the bandwidth you would need to adjust the byte
counts like:
(1500 * 8)/92 = 130 bytes.
But 130 bytes is rather small - so let's use 15,000 and 1300 bytes.
Sanity check, 1300/15000 = 8.6%
(I think this would be reasonable given a 1.544Mb/s line has 193,000 bytes
to send each second.)
So setup queues like this:
interface s 0
custom-queue-list 1
!
!
queue-list 1 queue 1 limit 100
queue-list 1 queue 1 byte-count 1300
queue-list 1 protocol ip 1 tcp 80
queue-list 1 queue 2 limit 100
queue-list 1 queue 2 byte-count 15,000
queue-list 1 default 2
At 10:20 AM 12/28/2000 -0500, you wrote:
>Hi,
>
>Is anyone aware of a URL that describes how figures used for the configuration
>of traffic shaping are derived? It seems the ones I have found start somewhere
>in the of the problem resolution.
>
>Particularly the relationships between time-period, be, bc.
>
>A site for instance, that describes from the outset how to determine the
>configurations for the the following problem...If over a 1.54 mbps frame
>connection, http traffic was to be limited to 128K. How would burst rates and
>time-period determined?
>
>Are the options to drop excess or reclassify with tos? How would becn and fecn
>be applied in a most fundamental way?
>
>Thanks in advance,
>
>Curtis
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:26:11 GMT-3