Re: (ASCEND) TNT HDLC/modem resource allocation

[Scott Kozicki <scott.kozicki@bluestar.net>]

Peter,

I know this doesn't answer your question directly, but it is somewhat
related. There are serious issues with the algorhythm that Ascend uses to
allocate HDLC resources. 

We came across ours when you attempt to do nailed T1's on a DS3 hub using
the channelized DS3 card for the TNT. There are no Spider chips on the CT3
card, so it uses HDLC resources from an HA-192 card (the HDLC card). Since
there are only 6 Spider chips on the card, each handling a total of 30 HDLC
(read: DS0) resources, then you can only have 6 nailed T1's up on the CT3
card. Kinda stupid.

What makes it worse is that if there's an analog/ISDN call that comes in
and takes the HDLC resources on a chip, then that chip is passed over for
nailed connections. What happens is that if you happen to have a nailed
connection fail for whatever reason (pipeline losing power?) and the TNT
tries to reestablish it, the algorhythm for HDLC allocation will start
cycling through available HDLC resources and gobble them up. It siezes the
HDLC's faster than it releases, hence you get a racing condition and end up
with a box that has *no* HDLC resources, despite having 3 HDLC-2 cards in
it. Connection profiles will show as up, but no packets will pass, giving
you a red herring to find the problem.

Think you're safe using a Frame Line 10 card, which has it's own HDLC
resources on it (just like your modem card)??? Think again. The algorhythm
basically chews up resources throughout the box, but doesn't ever report
what's actually going on. We had down circuits that were coming in on FL10
cards because of this problem, even though there were *PLENTY* of free HDLC
resources throughout the box. The only solution was to bounce the box,
bring up connections step by step, and use call routing to make sure PRI
based calls were hitting specific HDLC resources and nailed connections
used others.

A nightmare to say the least.

I wanted to mention this to the list because if you're thinking of using
DS3 hubbing to save some money - DON'T USE THE CT3 CARD! Buy an M13 box and
break the T1's down onto FL10 cards. You'll still prolly need an HDLC card
for PRI based calls, and to have some 'spare' resources just in case the
box gets heavily utilized.

In answer to Peter's query about call routing, I think the default is to
always utilize the HDLC's onboard before even hitting the call routing
logic. So Ascend might be right. The other problem here is that HDLC-1
cards have less Spider chips on them than HDLC-2 cards, hence less HDLC
resources to go around. This might be why you're still running out of modem
ports even though you're using the HDLC card as overflow. The bottom line
being that all HDLC resources on Ascend's gear are woefully anemic and the
process of allocation is broken.

Hope this helps,


Scott Kozicki
BlueStar Communications
++ Ascend Users Mailing List ++
To unsubscribe:	send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd:	<http://www.nealis.net/ascend/faq>