[comp.dcom.sys.cisco] Switched T1 with ciscos?

tjs@uh.msc.umn.edu (Tim Salo) (07/24/90)

Is anyone using (pseudo) switched T1 circuits, such as Accunet Reserve,
with cisco routers?

Our particular application has a 56k circuit between two locations.  At
times, more bandwidth is needed.  One possibility is to have an Accunet
Reserve line in parallel with the 56k circuit.  The Accunet Reserve T1
circuit would be enabled when additional bandwidth is needed.

   !--------!  56k line (up all the time)         !--------!
   !        !-------------------------------------!        !
   ! router !  switched T1 (up only when needed)  !        !
   !        !-------------------------------------!        !
   !--------!                                     !--------!

Has anyone tried this?

Will IGRP handle the addition and deletion of the T1 line gracefully?  (Will
the 56k line be used at all when the T1 line is active?)  ((To digress for
a moment, if the T1 line has a significantly different propagation delay
than does the 56k line, odd things can happen, e.g., large reassembly
queues.))

Tim Salo
tjs@msc.edu
(612) 626-0347

leinwand@hpccc.HP.COM (Allan Leinwand) (07/26/90)

> / hpccc:comp.dcom.sys.cisco / tjs@uh.msc.umn.edu (Tim Salo) /  3:47 pm  Jul 23, 1990 /
>    !--------!  56k line (up all the time)         !--------!
>    !        !-------------------------------------!        !
>    ! router !  switched T1 (up only when needed)  !        !
>    !        !-------------------------------------!        !
>    !--------!                                     !--------!
> 
> Has anyone tried this?

Yes.  Here at HP we have code that given the above situation will 'turn up' 
the interface on the router as needed.  Our setup involves a 56Kb line
which is always in use and a Switched 56 circuit which is only called
when needed.  Basically the setup requires one interface on the router
administratively down with the other active.  When the code decides to
turn on the downed interface, this raises DTR and the Switched 56 modem
automatically dials the call.

We dial the Switched 56 call when we see a high error rate
on the original 56Kb, or perhaps the utilization exceeds a user defined
threshold.  The code allows the user select the error rate and/or utilization
threshold which must be seen on the original circuit before the Switched 
56 line takes affect.  The code also sends mail informing the network 
operators about the new circuit in place and the reason it was activated.

In guessing the next question:  I don't know if I can give the code out.  Right
now it works *real* good on HP-UX :-)  If there is significant interest, 
perhaps we can arrange something.

> 
> Will IGRP handle the addition and deletion of the T1 line gracefully?  (Will
> the 56k line be used at all when the T1 line is active?)

I would think that the 56Kb in your situation will not be used unless
load on the T-1 becomes significant.  Also, it depends on the minimum 
bandwidth on the path to the destination.  

In our scenario the additional 56Kb (or whatever speed) is handled by IGRP
nicely.

Thanks,

Allan Leinwand
Hewlett-Packard
Corporate Telecommunications
Network Solutions Engineering
(415)-424-3738
allan@hpcta.corp.hp.com

"I'm going to hell in a bucket, but at least I'm enjoying the ride"

rob@oc.rjl.com (Rob Liebschutz) (07/27/90)

> Will IGRP handle the addition and deletion of the T1 line gracefully?  (Will
> the 56k line be used at all when the T1 line is active?)  ((To digress for
> a moment, if the T1 line has a significantly different propagation delay
> than does the 56k line, odd things can happen, e.g., large reassembly
> queues.))

When the T1 line comes up, the additional bandwidth of the 56kb line
becomes a small enough drop in the bucket that its hardly worth using
and will most likely degrade your performance due to the higher delay.
By default, IGRP will route all of the traffic over the T1 link, but
the "variance" configuration command can be used to override this in a
probably unsuccessful attempt to load balance between the two lines.
Load balancing does in fact work very effectively for two lines with
similar bandwidths.

Rob