From: Jason T. Rohm (jtrohm@xxxxxxxxxxx)
Date: Mon Oct 02 2000 - 16:14:19 GMT-3
I had another thought on this one... since the IOS uses the bandwidth
statement in the fragmentation algorithm, what happens when the line speed
does not accurately reflect the speed? ie. I put 28K in for the bandwidth,
but the line is crappy and only connects at 14.4K... wouldn't this cause
packets to build up on the analog side? MLPPP would resequence them, but I
can the the potential for all kinds of timing issues here...
As an aside to this... is there a way to make the bandwidth on the interface
change on each dial-up?
-Jason
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Kenneth Sacca
Sent: Monday, October 02, 2000 11:03 AM
To: Jason T. Rohm; ccielab@groupstudy.com
Subject: Re: MLPPP frame dispatch/speed
Jason,
I believe we use method two. You would use the following command
interface virtual-template1
bandwidth 78
ip address 131.180.60.5 255.255.255.0
ip mroute-cache
ppp multilink
ppp multilink fragment-delay 8
ppp multilink interleave
Fragment size is calculated using the following method.
fragment size = bandwidth X fragment-delay /8
So based on my example above, the fragment size would
be set to 78 bytes.
Hope this helps
Ken
"Jason T. Rohm" wrote:
>
> Does anyone know where I can find documention on Cisco's implementation
of
> RFC 1990 (MLPPP)? In particular I am interested in the way Cisco has
> implemented the fragmentation and dispatch mechanisms.
>
> I have a 1601 w/BRI WIC here. I am trying to determine if adding an
analog
> dial-up to the MLPPP bundle will increase performance. The reason I am
> questioning this is because all of Cisco's documentation says that the
> fragements are dispatched in a round-robin fashion.
>
> This would mean that the actual bandwidth available might be (Nx) where x
> is the lowest bandwidth the channels (3 x 33.6 = 100.8 kbps). (See RFC
> comments below.)
>
> The RFC outlines two methods of breaking up the packet. First, break up
the
> packet into unequal fragments using the bandwidth of the channel to
> determine size. Second, break up the packet into numerous small fragments
> and dispatch the fragments proportionately to the bandwidth.
>
> If Cisco uses method one, then a round-robin dispatch would make sense.
> However, if this is true, how does the IOS determine the bandwidth of the
> analog line? (Since the external modem does not return a connect speed).
>
> Thanks
>
> -Jason
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:25:23 GMT-3