NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/bit/listserv/biglan/67

Text File | 1992-08-17 | 9.2 KB | 193 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!paladin.american.edu!auvm!SUVM.BITNET!BIG-REQ
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.big-lan
Approved: [e-mail]
Date:         Tue, 18 Aug 1992 08:25:07 EDT
Sender:       Campus-Size LAN Discussion Group <[e-mail]>
From:         [e-mail]
Subject:      BIG-LAN Digest, Volume 4, Number 54, Tuesday, August 18, 1992
Lines: 181

BIG-LAN DIGEST           Tuesday, 18 August 1992       Volume 4 : Issue 54

Today's Topics:

                    Re: "evil" nature of hubs/hublets
                           "two repeater rule"

Moderated by John Wobus, Syracuse University

Relevant addresses:
                                  Internet                BITNET
Submissions:              [e-mail]     BIG-LAN@SUVM
Subscriptions:    [e-mail]     BIG-REQ@SUVM
LISTSERV/Archives:       [e-mail]    LISTSERV@SUVM
Moderator:                [e-mail]              JMWOBUS@SYREDU
Anonymous ftp archives:           syr.edu

Note: BIG-LAN is redistributed through many mailing lists at other sites
run by other individuals.  If you subscribe(d) through such a
"redistribution" list, you will need to remember its owner.

syr.edu also has a copy of the BIG-LAN "FAQ" memo (answers to frequently
asked questions) under the path information/big-lan/big-lan.faq

----------------------------------------------------------------------

Date:        Mon, 17 Aug 92 16:13:28 BST
From:        Richard Letts <[e-mail]>
Subject:     Re: "evil" nature of hubs/hublets

[e-mail] writes....
>
> From: "John M. Wobus" <[e-mail]>
> Subject:      Re: "evil" nature of hubs/hublets
>
> I mailed it out the first two times).  Here are the rules: there can be
> no more than 4 repeaters and 3 coax segments between any two stations
> on an Ethernet.  By "coax segment", I mean thickwire or thinwire.
> The other segments can be fiber or twisted pair.
>
Nope, these aren't the rules:
 -     You can have 3 data segments and 2 link segments between the
    stations. (including the segments the stations are attached to)
 -    Stations can only be connected to the 'data' segments
 -    The segments can be of any type.

Richard
- --
Richard Letts            My opinions are my own, and do not necessarily
Network Manager                 reflect those of my employer
University of Salford               Email: [e-mail]
United Kingdom                Phone : [phone]
United States of Europe

-------------------------------------------------------

Date: Mon, 17 Aug 92 12:29:13 -0400
From: William H. Magill <[e-mail]>
Subject:      "two repeater rule"

> >    I suspect that the real problem is that you are violating the 2 repeater
> >    rule. Both a "Hub" and a "hublet" are repeaters. Hence if you put
> >    one "hublet" on a "hub" you are ok. but as soon as you put the second
> >    "hublet" on you are probably in "theoretical violation" of the rule.
> >
> >I don't believe this is the case. ...
>
> I pulled the "I don't believe" out of context, but I want to make it
> clear that adding a repeater may or may not extend you too far.  It
> depends upon what's already in the network.  I made sure this was
> documented in the latest BIG-LAN FAQ (though that part was cut off when
> I mailed it out the first two times).  Here are the rules: there can be
> no more than 4 repeaters and 3 coax segments between any two stations
> on an Ethernet.  By "coax segment", I mean thickwire or thinwire.
> The other segments can be fiber or twisted pair.
>
> The point is: add a hublet only if you aren't already 4 repeaters away
> from some station on the ethernet.
>
The real issue in ANY CSMA/CD Carrier Sense, Multiple Access/Collision
Detection - network ie. Ethernet IEEE 802.3 - is the ability for the two
FURTHEST stations on a segment to "simultaneously" detect a collision,
back-off and to retransmit. This dimension is measured in TIME, not feet.
The FURTHEST separated stations are NOT necessarily the two stations which
you are interested in. But they represent the LONGEST POSSIBLE path in TIME
through that physical segment.

To quote from the spec: 4.1.22 Access Interference and Recovery
"If multiple stations attempt to transmit at the same time, it is possible
for them to interfere with each others transmissions, in spite of their
attempts to avoid this by deferring. When transmissions from two stations
overlap, the resulting contention is called a collision. A given station
can experience a collision during the initial part of its transmission (the
collision window) before its transmitted signal has had time to propagate
to all stations on the CSMA/CD medium. Once the collision window has
passed, a transmitting station is said to have acquired the medium;
subsequent collisions are avoided since all other (properly functioning)
stations can be assumed to have noticed the signal (by way of carrier
sense) and to be deferring to it. The time to acquire the medium is thus
based on the round-trip propagation of the physical layer whose elements
include the PLS, PMA and physical medium."

PLS = Physical layer signaling
PMA = Physical medium attachment

What all this is saying is that all of the "electrical characteristics of
signal propagation" - voltage level, wave form shape, capacitance, etc. -
control the whole ball game. This is why a "Fat Orange" Ethernet cable
with a 90 degree kink in it becomes a major problem. Not because the braid
and the conductor actually short out, but because the distortion effects
the total signal propagation.

Because there is ONLY ONE way to determine the REAL characteristics of any
given segment - and that is with expensive and extensive instrumentation
(the minimum tool needed is a TDR) the original DIX crowd (Digital, Intel
Xerox) formulated a series of "rules and budgets" that were "conservative"
in their goal to achieve Interoperability. Consequently we find that a
Segment can be so many meters in length if it is Fat Orange, or Thinwire,
or Twisted Pair. Similarly, the individual components - taps, repeaters,
connectors, etc were all given "budgets" based in feet to be subtracted
from these maximums when designing a network.

The truth of the matter is that these numbers are "conservative" consistent
with "sound installations." It is very true that in any given
configuration, with any mix of hardware and at any given point in time, one
can violate these "rules and budgets" with impunity. This becomes
especially true with today's "faster" hardware. While the propagation delay
of a "real" DEC DELNI might have been 10 feet back in 198? today, an
equivalent 0>piece of equipment might be worth 2 feet. Similarly, any given
vendor can give you very exact figures and rules for their own hardware in
a homogeneous environment which far exceed the "Standard rules and
budgets." But when asked to provide those same assurances in a mixed vendor
environment, you will find that they will only "certify" the numbers
mentioned in the standard.

You will also find that the physical topology of the network may NOT effect
the logical use of the network. It is entirely reasonable that you might
have a single segment with 3 distinct and essentially non-overlapping areas
of logical interconnection - both ends and the middle. It is quite possible
that these three groups might be able to conduct "reasonable" business on
the same segment even though collectively they are violating the rules and
impacting throughput.

As they say, your mileage may vary.

Some of the most evil things in this area are now surfacing in the 10BaseT
world. Because the PC folks think that they invented radial wiring the are
more than happy to ignore the experiences of Thinwire. The MOST EVIL thing
which constantly occurs in a 10baseT wiring setup is "telephone-itis."
It gets installed to look just like a telephone setup. But even
"contemporary" telephone practice is ignored. (The Telephone world has
discovered that Digital Switches are VERY unforgiving about wiring
practices which worked for 50 years and more under POTS - plain old
telephone service.) Since twisted pair "punches down" quite nicely on a
"157 Block" in the wiring closet, they get used a lot. A 157 Block has a 25
pair "amp" connector on the side making it very easy to "cross connect"
two sets of things. The problem comes in that the 10BaseT spec only allows
1) a very short run 2) a limited number of splices and 3) very specific
"twist" statements. So you find that two 157 blocks 4 feet apart, get
interconnected by a "standard" 25 foot long Telco 25 pair cable.
Not only do you suddenly have an extra 21 feet in your cable run, but you
also have added 4 splices where they are not only unnecessary, but
detrimental.

You wind up with:
wallplate(cable)157block/J-connector/J-connector(cable)J-connector/J-connector/
RJ45block/RJ45(cable)RJ45[hub].

Instead of:
wallplate(cable)RJ45block/RJ45(cable)RJ45[hub].

William H. Magill                         Manager, PennNet Computing Services
Data Communications and Computing Services (DCCS)  University of Pennsylvania
Internet: [e-mail]                   [e-mail]
          [e-mail]

-------------------------------------------------------


End of BIG-LAN Digest
*********************