jaap@mcvax.UUCP (Jaap Akkerhuis) (01/23/85)
According to the Ethernet specification Versions 1.0 and 2.0:
"A maximum of two repeaters (or four half- repeaters, i.e. two remote
repeaters) may be in the signal path between any two transceivers on
the channel"
Ethernet Specification Version 2.0, November 1982, Paragraph 7.6.4,
page 83.
To my suprise I'm offered two different brands of repeaters, which claim they
can have more then two.
Sension claims a maximum of six, and Isolan (Manufacturer BICC data networks
LTD) 4.
I don't have (yet) the info about Sension, but to quote the Isolan sales
blurb:
"It performs signal retiming and communication between the two segments
(...). Longer distance links between segments up to 2km (!) apart can
be implemented width two (not two half-repeaters?) repeaters and fibre
optic transceivers.
Because each repeater fully regenerates each packet including
preamble, up to 5 segments may be connected in series, allowing a
network to be extended to 3km in diameter. The repeater performs the
functions of regenerating preamble, extending collision fragments,
enforcing the inter-packet gap, and automatic partitioning of the
netwerk in the event of a segment failure. (...) The repeater has been
implemented with the minimum of components (...) "
The box also claims to have been designed to the following list of standards:
ISO DIS 8802/3, IEC 380, IEC 435, IEEE 802.3, FCC 20780, ECMA 80 and 81,
ECMA TR/19, VDE 0871, VDE 0875 level N.
(Everything in the parenthesis are my comments, or left out non-interesting
details).
Well, notice the contradiction in the statement.
Anyway, the big question is of course "Can this be true? Has anyone experience
with the named products?"
Both firms are based in Engeland so another question arises immediately:
Could this be another feed to the famous quote "It's British so it
probably doesn't work".
Please answer by mail and I summarize to the net.
Jaap Akkerhuis, {seismo,decvax}!mcvax!jaaprpw3@redwood.UUCP (Rob Warnock) (01/31/85)
+--------------- | To my suprise I'm offered two different brands of repeaters, which claim they | can have more then two... | Sension claims a maximum of six, and Isolan (Manufacturer BICC data networks | LTD) 4. | Anyway, the big question is of course "Can this be true? Has anyone experience | with the named products?" +--------------- (Ah... the stuff consultants thrive on! "Fee, Fie, Foe, Fum, I smell the blood\\\\\ money of a naive one..." ;-} ) What we have here is a little "version skew" in the spec, methinks. While it is true that the Ethernet 2.0 spec still says a maximum of two repeaters across the "diameter" of the net, that limit seems to be a holdover [read: "bug"] from the 1.0 spec which did NOT require that repeaters regenerate the preamble back to 64 bits. In 1.0, two repeaters was all you could have, because the various allowed "bit eatings" (bits lost at the beginning of a packet) of transceivers and coder/decoders along the path wouldn't leave enough preamble bits if you had more than two repeaters. But 2.0, like 802.3, requires repeaters to regenerate a full 64 bits of preamble. (Compare section 7.6.4.1 in both E/1.0 and E/2.0). Therefore, the real limit on repeaters is the round-trip propagation delay limit. There is a certain allowed "start up" delay of a transceiver, a PLL, a decoder, an encoder, and a transceiver. Each repeater would thus shorten the permitted physical "diameter" of the network, to ensure worst-case end-to-end prop delay of 46.4 usec. Unfortunately, this makes the whole issue configuration dependent, as less cable means more allowed repeaters. (If I recall, 802.3 just said "maximum of four repeaters".) I expect that the only way people are going to keep it straight is if repeater manufacturers label their products with the "effective cable length", so you can subtract the sum of your repeaters delays (less the two repeaters "included" in the standard) from 2500 meters to get the allowed "diameter". So assuming your repeaters are 2.0 conforming, and the distance was not too great, you COULD use two "full" repeaters and a piece of Ethernet cable (or fibernet) to make one "pair of half-repeaters with included point-to-point link". (Good luck. Let me know if it works.) Rob Warnock Systems Architecture Consultant UUCP: {ihnp4,ucbvax!dual}!fortune!redwood!rpw3 DDD: (415)572-2607 USPS: 510 Trinidad Lane, Foster City, CA 94404
dw@rocksvax.UUCP (Don Wegeng) (02/01/85)
In article <425@mcvax.UUCP> jaap@mcvax.UUCP (Jaap Akkerhuis) writes: > >According to the Ethernet specification Versions 1.0 and 2.0: > > "A maximum of two repeaters (or four half- repeaters, i.e. two remote > repeaters) may be in the signal path between any two transceivers on > the channel" The reason that you are not suppose to use more than two repeaters is that many of them drop up to 16 bits from the preamble while synchron- izing to the timing, etc (you cannot predict the exact frequency of the transmitter). Therefore, if you had three repeaters in the path you could possibly lose the entire preamble, therefore leaving nothing for the receiver to sync it's timing from before the *real* data starts. I suppose that somebody may manufacture a repeater that doesn't drop as many bits. If so then you ought to be able to connect more of them into the net. However, note that this would be in violation of the Ethernet Version 2.0 spec (see page 83 for more details). You'd have to do some math to calculate the maximum length that a net can be. -- /Don "Don't touch me, I'm a real live wire. Psycho Killer..." arpa: Wegeng.Henr@Xerox.ARPA uucp: {allegra,princeton,decvax!rochester,amd,sunybcs}!rocksvax!dw || ihnp4!tropix!ritcv!rocksvax!dw
jqj@cornell.UUCP (J Q Johnson) (02/01/85)
I don't have any information on other repeater vendors, but I gather
that the DEC Ethernet repeater is advertised NOT to eat any preamble
bits. It does, of course, contribute to the propagation delay budget
as the Ethernet spec. allows. This would seem to imply that an
Ethernet using only DEC repeaters could comfortably have more than
2 repeaters between a pair of nodes if the maximum cable distance
between nodes was suitably reduced. For example, consider the
following topology which violates the spec in various ways:
( 500m Ethernet backbone)
----------------------------------------------------------------
R R R
|(500m seg) |building 2 (500m seg)|
building 1 | building 3
(500m seg)| building 4
________/R---------------------
The fact that the repeater for building 2 is in the middle of the
backbone Ethernet would seem to imply that the delay budget for
the repeater connecting to building 4 can be safely ignored.
Of course, in practice you would have to figure the delay budget
carefully to make sure everything worked. The advantage of the
Ethernet spec as written is partially that it is fairly simple --
you don't have to calculate delay budgets between every possible
pair of nodes. Since configuring an Ethernet is already a complex
undertaking, making it still more complex by permitting some but
not all configurations of multiple repeaters seems like a dangerous
idea.