[c-nsp] Multilink PPP vs. pinhole congestion/DI vs. VWIC2-MFT*

[Andre Beck]

Hi,

according to what I've read here (not that much to find on CCO) the
IOS implementation of MP (also called MLPPP or MLP hereabouts) is
automatically applying weights to bundle members of different
bandwidth and, if fragmentation is not disabled, will derive the
per-member fragment size from this weight. This would allow to
avoid pinhole congestion in such cases. What I'm interested in is
the actual experience with such setups. I'm planning to bundle
two or three E1s, where one of these will have several timeslots
stripped from it for the sakes of drop and insert. The easiest
way to handle this would be a VWIC-2MFT-E1-DI and relying on the
unequal weight fragmentation capabilities of IOS's MP. The E1 in
question will probably have a maximum of 10 timeslots less than
the other(s), so the remaining bandwidth would be better than 60%
compared to the others. Any drawbacks to expect here?

BTW, anyone already tried the all-new shiny VWIC2-?MFT-T1/E1 and/or
-G703 VWICs? They are supposed to have DI included and in 28xx/38xx
chassis even allow chassis wide DI crossconnects(!). Ideally, using
these VWICs I could distribute my, say, 9 B-channels to 3 channels
each on 3 E1s, leaving symmetric 28 slots for data[1]. The basic
question here is: Is it worth the hassle? Or will the fragmentation
weight just do the trick? If I read the blurb about the VWIC2s correctly,
they will not allow this inter-VWIC TDM switching in just the chassis
WIC slots of a 28xx but require special NMs, right?

TIA,
Andre.

[1] I have basically found *no* description of what is possible here
    and what isn't. It is just an assumption that using two of these
    VWIC2-2MFT-T1/E1 I could

    a) Connect E1#0 to a PBX and have channels 1-9 active on it
    b) xconnect channel 1-3 from E1#0 to channel 1-3 on the E1#1
    c) xconnect channel 4-6 from E1#0 to channel 1-3 on the E1#2
    d) xconnect channel 7-9 from E1#0 to channel 1-3 on the E1#3 and
    e) plug a "channel-group 0 timeslots 4-31" on E1#1-3

    so I can finally bundle the three resulting serials using MP.
    Anyone knowing anything more about this issue is welcome to
    point me into the right direction, even if it is the direction
    of the trashbin of ideas.
-- 
                  The _S_anta _C_laus _O_peration
  or "how to turn a complete illusion into a neverending money source"

-> Andre Beck    +++ ABP-RIPE +++    IBH Prof. Dr. Horn GmbH, Dresden <-