Ascend Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: (ASCEND) Bonded T1's?

On Tue, Dec 30, 1997 at 09:05:51AM -0800, Matt Holdrege wrote:
> At 10:52 AM 12/30/97 -0600, Scott Kozicki wrote:
> >Can you bond multiple T1 interfaces together on a Max 4K? That is to
> >say, can I MPP 2 T1's together to form one ~3MB pipe upstream?
> 
> Well, you can bundle multiple channels across T1's, but you can't bundle
> two unchannelized T1's together.

And I never understood why. I've posted a somewhat long pamphlet here
(Subject: The Nailing Game) that asked a lot of questions regarding
the partially strange handling of nailed connections in Ascend boxes.
I haven't seen much response on it yet (especially not from Ascend).
IMHO these questions need an answer: I can't understand why I need
Nailed/Mpp to high-level-bundle two nailed groups. An MP or MPP profile
would suffice. I don't understand why it is not simply this way:

1) There are groups. You can either apply a group to one B-channel or
   to a set of B-channels that are technically ready for low-level-bundling
   (like those in the same T1/E1 PRI or BRI). Any such group constitutes
   a - lets say - "channel". The serial WAN port has one group because
   it is another "channel". For channels that consist of multiple low
   level bundled B-channels a _complete_ documentation of how the bundle
   is actually built is necessary for PRI and BRI (I expect simple time
   slot incremental repeating but thats not documented anywhere).

2) You can high-level-bundle any channels. F.i. you could bundle two
   channels that each consist of low-level-bundled B's, could bundle serial
   WAN with two low-level-bundled B's of a BRI or whatever - it simply
   should work. And it has to work with MP or MPP in nailed mode. I
   don't see any need for Nailed/Mpp yet.

3) You can supply any such high-level-bundle you made according to 2)
   with a new facility called "add more bandwidth (for whatever reason,
   failover or DBA) by _dialing_ B-channels" by using MPP (for DBA, else
   MP would suffice either, but I wont nitpick here) and the Call Type
   Nailed/Mpp. That's what Nailed/Mpp is documented for and that's what
   makes sense.

Now the reality check: The above is how it should be, but some of the
components don't work. 1) actually works nice except documentation is
missing. 2) doesn't work at all - I cannot list more than one group
number in a MP or MPP profile when Call Type=Nailed. 3) works somewhat,
but a lot of statistical errors in detail make it almost useless. The
only case where it works nearly flawless is when bundling dialed B's to
a nailed B or a _high_ level bundle of nailed B's.

(This was a short version, the old post was much more detailed).

> However, beware that vanilla TCP/IP doesn't work well at all when you
> exceed about 6 to 8 MP channels. 

I think we can live with that - as soon as we can mix low-level and
high-level bundling without problems ;->

Happy New Year,
Andre.
-- 

Kanther-Line: PGP SSH IDEA MD5 GOST RIPE-MD160 3DES RSA FEAL32 RC4

+-o-+--------------------------------------------------------+-o-+
| o |               \\\- Brain Inside -///                   | o |
| o |                   ^^^^^^^^^^^^^^                       | o |
| o | Andre' Beck (ABPSoft) beck@ibh-dd.de XLink PoP Dresden | o |
+-o-+--------------------------------------------------------+-o-+
++ Ascend Users Mailing List ++
To unsubscribe:	send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd:	<http://www.nealis.net/ascend/faq>
References: