Re: [SLUG] [OT] repeaters & hubs ##

[Terry Collins <terryc@nospam.woa.com.au>]

Ben Donohue wrote:
> 
> DaZZa wrote:
> 
> >
> >
> > You can extend your network using hubs as well - as long as you apply the
> > 5-4-3 rule to the total segment length and number of repeaters.
> 
> Okay, what's the 5-4-3 rule?

From the archives

.............................
> 
 ] >  Of course - I've *never* seen a network set up like this.  Heck,
I've only met two
 ] >  people who've ever heard of the 5-4-3-2-1 rule in my life.  
 ] 
 ] </p.hanson> Please explain? <\p.hanson>

 Look you complete and utter .. oh, hang on.  Perhaps if I try to be
polite *and*
 informative, a) you'll believe me, and b) the world will be a better
place.  I'll
 give it a burl, and see what happens.

 And shouldn't that be <p.hanson> text </p.hanson>  ?   ;)

 Anyways - I just tried to find some documents I could point you to on
the web, but
 alas these days the ieee.org page seems to be designed with two
purposes in
 mind - firstly to prevent people finding information they want (for
free), and secondly
 to encourage people to pay for information they don't want.

 So, this is from memory.  Doubtless I'll be corrected if I present
any misinformation.

 The 5-4-3-2-1 rule means you can have 5 segments, 4 repeaters, 3
segments that
 are active, 2 that aren't, and that's all on 1 network.  Or something
close to that . . .

 The following qualifications / explanations apply (keeping in mind
that pretty much
 every term in networking has been appropriated over the years to mean
a variety
 of usually self-conflicting things) :

 Segment - all the wire(s) between hubs.
 Repeaters - electrical, hubs, call them what you will - they are
non-intelligent
    devices - they merely repeat (and perhaps boost) the electrical
signals down
    the wire.  They do not examine the contents of a packet, at any
level.
 3 segments that are active - these are where you can put DTE's (for
want of
    better description .. uhm.. hosts.. PC's.. etc) onto the 'overall
network'.
 2 segments that aren't - these are hub <---> hub conections that
can't have
    any DTE's on them.
 1 network - collision domain, in the CSMA/CD sense <npi>.

 As I say - I don't think I've ever seen a network designed like
this.  Let alone
 with the *really* tricky stuff - like working out the maximum
distance between
 any two DTE's on one collision domain.  All the specs are, of course,
precisely
 defined. OTOH, there's never a good reason to go beyond spec with
this
 stuff - particularly since trying to trouble-shoot it later is a
fnooting nightmare.

 If you want to try your luck at finding more on this, search down the
802.3
 specification.

 ] I was being a bit purist here. We effectively used the thick net at
the
 ] install fest how you suggest. For a network with equal priority
boxes, like
 ] home, this would be fine. Unfortunately, thin coax is not longer
 ] fashionable, nor up to some speed requirements of some business.

 {shudder.thicknet}  You really mean thicknet, as in the yellow
 I'll-buggered-if-I-can-bend-it stuff?  Ergh.

 I don't understand the meaning of 'equal priority boxes' (?). 
Anything on
 ethernet (in the CSMA/CD sense, not the utp, coax, or thicknet sense)
 is working in an 'equal priority' environment.  

 ] For a hierachial network, I'd chain the hubs with high priority
gear off
 ] the top hub. Of course it would be a switching hub anyway and we
would
 ] probably be talking 100MB uplinks anyway. 

     [ and ]

 ] A - It is better to have a flatter network, hence cascading is to
be
 ] avoided.

 {Confused again.}  Hierarchical network?  This sounds more like SDLC
 than CSMA/CD !  </nerdjoke>    A switching hub is an entirely
different
 kettle of fish.  It's effectively a bridgey-routery-kind of box.  It
far transcends
 the functionality of what you'd ( I'd ?) call a hub or repeater.

 And note that you're talking about an increase in speed of your
backbone,
 not an increase in *priority* of any particular device on the
network.  

 ] (Telesales at one site had three X 24 hubs daisy chaining off each
with one
 ] 10Mb UTP feed and continual slow response complaints. Deadly for
the NetPc
 ] they were considering. The obvious fix was to use the idle 2nd UTP
cable to
 ] split this load, but this didn't fix the problem noticably. Put in
Bay
 ] 100's with thier led load levels and it was like christmas trees
lights.

 I'm assuming that those hubs were saturated (?).  Still, that's only
70 (max)
 devices on a single collision domain.  Something I'd expect a segment
 (oooh, there I go misusing that word myself) to handle easily.

 What were they running?  I've seen networks running more PC's than
that,
 with everything (bar DOS and network drivers) coming off a server, at
10Mbps
 all the way, and coping fine.

 Depending on the nature of their work, you may have gotten away with
 sticking in two low-end Pentium linux boxes with dual-NIC's in each,
 and splitting their network into three bits.  How much are those Bay
100's
 again..?  ;)

 Cheers,
 Jedd.
> 
> --
> SLUG - Sydney Linux Users Group Mailing List - http://www.slug.org.au
> To unsubscribe send email to slug-request@nospam.slug.org.au with
> unsubscribe in the text

--
   Terry Collins {:-)}}} Ph(02) 4627 2186 Fax(02) 4628 7861  
   email: terryc@nospam.woa.com.au  www: http://www.woa.com.au  
       or terryc@nospam.zip.net.au 
   WOA Computer Services <lan/wan, linux/unix, novell>
   snail:  PO Box 1047, Campbelltown, NSW 2560.

 "People without trees are like fish without clean water"
--
SLUG - Sydney Linux Users Group Mailing List - http://www.slug.org.au
To unsubscribe send email to slug-request@nospam.slug.org.au with
unsubscribe in the text