(somewhat lengthy, but addressing several people's questions...)

Cynthia Fowler wrote:
>
> We have a remote location at which we must have a DTC for connecting
> a serial printer. One of our software applications requires
> associating with the printer to print checks. Our HP979/400 uses a
> 100baseTX LAN. We have a 10baseT but it is unused.

Note to Gavin Scott, who asked on this thread:

GS> Sorry if I wasn't paying attention to the thread up to this point,
GS> but are we sure that you can run DTS over 100Mb interfaces on the
GS> 3000?  I have this vague recollection that the DTC software on the
GS> 3000 might only bind to a 10Mb link.

From the above, Cynthia is running her local DTCs off the 100TX LANIC.
It should work just fine given a switching hub/bridge somewhere in the
loop to handle the speed conversion.  But if everything was plugged into
a 100TX hub, it wouldn't work; DTCs can't speak 100TX no more
than they can speak to a frame relay link or serial line, but with a
device in the middle that converts media (in this case 10BT to 100TX)
it works just fine as Cynthia points out.

> Our local DTCs work fine in this environment. However, yesterday my
> CE was here to redo our OpenView DTC Manager map and configure in a
> DTC16MX at our remote location when he said he did not believe we
> could run this DTC at a remote location over 100baseTX. The remote
> location is well over 100 miles from the main office. At the remote
> location we have a router with AFCP bridged and a 100baseT Switch.

Aha, a router and a switch on both ends; that's doing the media
conversion.  The rest is layer 2 (or above with routed AFCP).  The
10/100 issue is strictly layer 1.

> Our CE was about 95% sure that this cannot be done, but is checking
> to confirm or deny his suspicions.

It should work just fine (given adequate WAN bandwidth).

> Has anyone on this list ever attempted to do connect a DTC at a
> remote site over 100baseTX and been successful?

Again, you are confusing layer 1 issues.  You're not going "over
100baseTX", you have intervening switches/routers/some WAN link in
between doing the layer 1 conversions.  Your layer 2 is "same old thing"
with bridged AFCP, or you skip it and go to layer 3 if you are using
routed AFCP capable DTCs.  To exaggerate the irrelevance of layer 1,
consider the home user connecting on an analog modem, ISDN line, cable
modem, or DSL to an ISP; this then goes over WAN links of varying speeds
and encapsulations (frame?  ATM?  DWDM?) to arrive at the destination
where it converts back to who-knows-what on the other end.  You only
have to retain layer-3 compatibility in this case (TCP)
and everything else is handled for you.  With non-routed DTCs you have
to have a layer-2 path (bridge or TCP-encapsulated tunnel) to the other
end.  But layer 1 certainly doesn't fit into the picture.

Dennis Heidner added:

> However, in my opinion it makes more sense to let the DTC's use the
> slower 10Mbs interface that is built on the MIO card, and use the
> 100BT interface for inbound telnet, ds, ftp, web, etc.

Truly ideal, and keep the MIO 10Mb "subnet" off the network if possible
so that they are a "closed system" and not competing for bandwidth with
TCP traffic.  Of course this cannot be "closed" if you have DTC Manager
and want it to see your real network (it would have to reside on the
private subnet), or if you have TACs (Telnet Access Cards) in the DTCs
which must be IP-addressable.

Tom Emerson asks:

> I realize this isn't the exact scenario previously described, but
> wouldn't going to a 10/100 switch [converter] allow for more
> "overlap" when dealing with multiple DTC's?  "in theory", since
> 100mb is 10 times faster than 10mb, you should be able to "service"
> 10 DTC's consolidated through a 10/100 "switch" and a 100mb
> interface on the HP in the "same amount of time" as a single DTC
> being serviced by a 10mb interface, or am I smoking something
> that can't be bought at the corner drugstore?

In an ideal world, yes; but in the real world, it is infested with
caveats :-)

If your DTCs are on 10Base2 (ThinLan coax) the media is half-duplex
shared 10Mb, you can't get more than 10Mbps/sec, period.

If your DTCs are on 10BaseT and connected to an "intelligent switch" you
get roughly 10x the throughput of the coax scenario.  You will never get
10Mbps of useful data over a 10Mbps link; you can get closer to 100Mbps
of useful data over a 100Mbps link if it's full duplex, but still not
100Mbps, so I prefer to talk in terms of throughput rather than raw
bitrate.  So it could be "10 times faster" if 10 DTCs were serviced at
the same time.

The "intelligence" of the switch varies greatly.  If you have a simple
hub that the DTCs are hooked to, it is no different from coax; the media
is shared and half-duplex.  Same as coax.  Next comes the "10/100
switching hubs" which essentially are a hub with a 100TX and a 10TX
bus.  During autonegotiation at power-on, the "switch" (I use that term
loosely) puts you on the 10Mb bus or the 100Mb bus.  The really dumb
ones don't even bridge the two together (they require two uplinks, one
for each bus) and yes, this is not a figment of my imagination, I have a
Bay 250 lying around somewhere I'd like to introduce to Mr Sledgehammer
(kudos to Allegro's Sportster page :-) ).

Slightly more intelligent "switching hubs" have a build-in bridge
between the two internal bus planes, but the fact remains that they are
still shared media and thus half duplex on both the 10Mb and 100Mb
sides, so again, you can't push 100Mb worth of host traffic to 10Mb
devices any faster than 10Mb at a time.

Real switches will permit this, given adequate diversity of 10Mb devices
served by 100Mb host and adequate buffers.  You can't sustain a 100Mb
stream to a 10Mb device (upper layers handle flow control through
handshaking, acks, retransmits, source quench, etc) but the buffers
"carry over" the load until upper layers catch up.

Jeff Kell <[log in to unmask]>