At 04:18 PM 6/11/01, Jianfeng Wang wrote:
>Actually, the switch diameter rule is to keep spanning-tree protocol
>working. Cisco switches sets their echo time to 2s and forward-delay to
>15s. In which case, diameter larger than 7 makes spantree failed to work.
Please do not send messages to me personally. If you are confident in your
comment, you will send it to the whole group and everyone can benefit. If
you aren't confident, then I don't want to hear it.
I checked IEEE 802.1D more carefully and it does say the recommended
maximum diameter of a bridged network is 7. There's no such thing as an
echo timer, but the hello timer is 2 seconds by default. The forward delay
is 15 seconds by default, as you say.
The explanation of how these relate to each other in IEEE 802.1D is
incomprehensible ;-), but Clark and Kennedy say that the 15 second forward
delay comes from calculating the end-to-end propagation delay on BPDUs,
with the following assumptions:
* BPDU hello timer is 2 seconds
* It takes a bridge 1 second to send a BPDU after receiving it on another
port.
This is called the bpdu_delay. (It's extremely conservative.)
* Up to 3 BPDUs may get lost as a message moves from one end of the network
to
the other.
* The maximum number of bridge hops between any two nodes is 7.
End-to-end BPDU propagation delay
= ((lost_messages +1) x hello_timer) +(bpdu_delay x (diameter -1)
= ((3 + 1) x2) + (1 x (7-1))
= 8 + 6 = 14 seconds
You also have to take into account a few other obscure things like the
transmit halt delay, the bridge transit delay, the max medium access delay,
the message age overestimate, the max frame lifetime, etc. It seems to be
extremely complicated. Argh.
Thanks
Priscilla
>J Wang
>
>Priscilla Oppenheimer wrote:
>>The 5-4-3 "rule" starts over with the switch. A switch port is the "demarc"
>>for a collision domain.
>>
>>Think of your network in hierarchical terms or in blocks. Design the shared
>>hub-based blocks using the 5-4-3 rule. Design the switched blocks with a
>>concern for scalability of the Spanning Tree Algorithm and broadcast
>>traffic. Switches forward broadcasts. Too many stations sending numerous
>>broadcasts in a switched network is a scalability issue. Cisco recommends
>>no more than a couple hundred stations be in a single broadcast domain.
>>
>>A "rule" for switches is that you shouldn't have more than 7 connected
>>together. But I don't think that's an actual rule you'll find in any IEEE
>>documents. However, it is a good idea to limit the size of a large switched
>>network to avoid the Spanning Tree Algorithm taking a long time to converge
>>and broadcast traffic becoming a concern. You can subdivide your switched
>>network with routers and VLANs. With VLANs, you can have one spanning tree
>>per VLAN.
>>
>>And a few more words from my basic Networking 101 "lecture," that I often
>>give to K-12 network engineers who are usually too busy to step back and
>>think about fundamentals!? ;-)
>>
>>A hub is just a way of making wiring less confusing. Instead of connecting
>>stations in a bus coax cable, we learned from the telephone industry that
>>we could use structured cabling, hubs and twisted-pair cabling. But that's
>>all a hub really does. It simplifies cabling (and amplifies the signal,
>>since it is a repeater.) A hub makes it possible for stations to share
>>bandwidth even though they are physically cabled in a star topology with an
>>active repeater (hub) between them.
>>
>>Although placed in topologies in the same place as hubs, switches are
>>completely different beasts. A switch has the opposite purpose of a hub. A
>>switch divides up the number of stations that share bandwidth. It reduces
>>the size of collision domains. A switch is a high-speed multi-port bridge.
>>A hub is a repeater.
>>
>>Priscilla
>>
>>At 12:05 PM 6/11/01, Hartnell, George wrote:
>> >A couple of 5-4-3 refinements:
>> >
>> >Try this picture:
>> > MORE HOSTS
>> > > \.../
>> > > hosts HUB(s) hosts
>> > > \|/ \|/ \|/
>> > >host --- hub --- repeat --- SWITCH --- repeat-- hub --- hub---- etc...
>> > > /|\ /|\ /|\
>> > > hosts hosts hosts
>> > >
>> >The question of 'slot time' becomes confusing to me with the addition of
>> >one, or more, switches into the (10,100,1000) circuit(s). The 'etc'
above
>> >asks, "What is the practical limit of cascaded switch/hub combinations
>> in a
>> >10 Mb/s ethernet? A combination 10 and 100 Mb/s ethernet? Yes, even
>> with a
>> >1Gb/s backbone?" Please, don't laugh, we here in K-12 education need to
>> >lash up all kinds of stuff. Sometimes, if you can believe it, people
will
>> >even sneak an extra hub into the building!
>> >
>> >No problem is insurmountable with the comparatively limitless private
>> sector
>> >resources. Network design is easy (easier) without the albatross of old
>> >equipment never, ever, being surplused out, just moved out closer to the
>> >edge. That's why we like the theory, as below. We frequently need to
>> figure
>> >out what flies ourselves.
>> >
>> >Very best, G.
>> >
>> >
>> >"Be strict in what you send, and forgiving in what you receive."
>> > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
>> > George Hartnell, Network Supervisor
>> > Bellingham School District, 1306 DuPont St.
>> > Bellingham, Wa. 98225-3198 (360)647-6860
>> > [EMAIL PROTECTED] Internet Mail
>> >
>> >
>> >
>> >-----Original Message-----
>> >From: Priscilla Oppenheimer
>> [mailto:[EMAIL PROTECTED]]
>> >Sent: Thursday, June 07, 2001 12:38 PM
>> >To: [EMAIL PROTECTED]
>> >Subject: Re: Rule 5-4-3 [7:7578]
>> >
>> >
>> >At 02:17 PM 6/7/01, Aleksey Loginov wrote:
>> > >Hi!
>> > >Question about rule 5-4-3.
>> > >How it's work for 10BaseT?
>> > >This scheme correct?
>> > >
>> > > hosts hosts hosts
>> > > \|/ \|/ \|/
>> > >host --- hub --- repeat --- hub --- repeat --- hub --- host
>> > > /|\ /|\ /|\
>> > > hosts hosts hosts
>> > >
>> >
>> >Great job on the ASCII art! It's quite pleasing to the eye. ;-)
>> >
>> >A 10BaseT hub is a repeater. Dare I say that there's no difference
between
>> >a Layer-1 repeater and a hub? Just like there's no difference between a
>> >Layer-2 switch and a bridge, or a Layer-3 switch and a router. Argh.
>> >
>> >One topology that became popular for explaining propagation delay on an
>> >Ethernet network was the 5-4-3 topology. With a 5-4-3 topology, you can
>> >have up to five segments in series, with up to four repeaters, and no
more
>> >than three mixing segments. If three mixing segments are used, then the
>> >remaining two segments must be link segments. A mixing segment is one
that
>> >can have more than one attachment, such as a coax cable. A link segment
is
>> >a point-to-point link. It could be a fiber link between hubs, for
example.
>> >A 10BaseT hub connection to an end station is also a link segment.
>> >
>> >The 5-4-3 topology is just one of many models that work actually. It is a
>> >simplification of the actual rule that says that the round-trip
>> propagation
>> >delay in one collision domain must not exceed the time it takes a
>> sender to
>> >transmit 512 bits, which is 51.2 5s for 10-Mbps Ethernet, and is also
>> known
>> >as the slot time.
>> >
>> >A single collision domain must be limited in size so that a station
>> sending
>> >a minimum-sized frame (64 bytes or 512 bits) can detect a collision and
>> jam
>> >signal reflecting back from the opposite side of the network, while the
>> >station is still sending the frame. Otherwise, the station would be
>> >finished sending and not listening for a collision, thus losing the
>> >efficiency of Ethernet to detect a collision and quickly retransmit the
>> >frame. For a higher layer to notice that the frame needs retransmitting
>> >takes much more time. Ethernet retransmissions occur within nanoseconds.
>> >
>> >Priscilla
>> >
>> >
>> > >Any comments, please.
>> > >
>> > >Best regards,
>> > >Aleksey
>> >________________________
>> >
>> >Priscilla Oppenheimer
>> >http://www.priscilla.com
>>________________________
>>
>>Priscilla Oppenheimer
>>http://www.priscilla.com
________________________
Priscilla Oppenheimer
http://www.priscilla.com
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=8069&t=7578
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
