sullivan@harvard.ARPA (John M. Sullivan) (08/01/85)
(I am posting this note for a friend. -- John Sullivan, sullivan@harvard)
I am about to ask the apocalyptic question, "How do I connect two
ethernets via microwave link?" so if you're not interested, now's the
time to skip the rest of this note.
I'm peripherally involved in a networking project that ultimately may
need to connect two distant ethernets by way of a microwave link.
Assuming that I have two bi-directional microwave towers pointed at
each other and running T1, I need some kind of magic to mediate
between each end of the T1 link and the respective ethernets.
Differences in transmission speeds are inevitably going to require
*some* kind of flow control. I am ignorant about the issues involved,
as will be demonstrated by the explicit and implicit questions in the
paragraphs that follow. I will be grateful for any suggestions,
advice, or referrals, from anyone.
Here are some ideas for the T1<->ether mediating magic. One possibility
I've heard about is a card from ACC which plugs into a VAX and transforms
T1 into something understandable. (Does anyone have any details on this
card?) However, this solution inevitably requires two hosts, each of which
will experience a hefty CPU burden.
Another possibility is some kind of host-based flow control scheme
using a couple of cheapie hosts (visuals, isi optimums, ibm pcs,
whatever). What kind of hardware and software would be required for
this solution?
Yet another idea (my favorite) is to use some kind of bridge device on
either end of the T1 link. Such a box would plug directly into the
arpanet, and squirt 10megabit ether packets to/from the T1 radios.
Apparently, Bridge manufactures a similar product (called a "bridge"),
but it translates ether to a protocol at 1Mhz (perhaps for twisted
pair?). Unfortunately, T1 runs at 1.544Mhz. Has anyone heard of a
T1<->ether bridge? The big advantages of this scheme are: no
software overhead, no host overhead, more-or-less off-the-shelf
hardware throughout the entire ether<->ether link, direct plug-in to
the ethernet.
Comments, other possible solutions, etc., are all welcome. I will
summarize whatever I get from the net and re-post them for those people
who express an interest in reading the results.
Please post letters directly to alan@harvard.{arpa,uucp,bitnet} or to
sullivan@harvard; unfortunately, I don't have direct access to net
news (my mail is forwarded).
My thanks to all those who waded through my verbiage, and my special
thanks to those of you who see fit to reply.
Al
-----
Alan Langerman
alan@harvard.{arpa,bitnet,uucp}
--
John M. Sullivan
sullivan@harvardron@brl-tgr.ARPA (Ron Natalie <ron>) (08/02/85)
> (I am posting this note for a friend. -- John Sullivan, sullivan@harvard) > > I am about to ask the apocalyptic question, "How do I connect two > ethernets via microwave link?" so if you're not interested, now's the > time to skip the rest of this note. > Use a Vitalink ethernet link. However it is only 228 KB not T1. It is a repackaging of the Bridge box. > Here are some ideas for the T1<->ether mediating magic. One possibility > I've heard about is a card from ACC which plugs into a VAX and transforms > T1 into something understandable. (Does anyone have any details on this > card?) However, this solution inevitably requires two hosts, each of which > will experience a hefty CPU burden. We are intending to use this card to link our team of local area nets at one site to similar remote sites. The system involves using high speed dedicated IP gateways (which currently are isi optimums, which are 68000 on Q bus systems, IBM-PC's would lose). > Yet another idea (my favorite) is to use some kind of bridge device on > either end of the T1 link. Such a box would plug directly into the > arpanet, and squirt 10megabit ether packets to/from the T1 radios. > Apparently, Bridge manufactures a similar product (called a "bridge"), > but it translates ether to a protocol at 1Mhz (perhaps for twisted The problem is that they flat out run a 68000 without any advanced I/O concepts like DMA or interrupts.