Desktop

Ask something
Discussion Locked

5, 4, 3 Rule

0 helpful
7 replies

by keith.abbott

Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

Ok Gentlepeople,

I have been having an ongoing debate regarding the practical meaning of networkings 5,4,3 rule, sometimes called the 5, 4, 3, 2 rule.

Imagine a 10/100B-T network with, for instance a server connected to a backbone switch (call it SW-A) which is in turn connected to another switch (B) which is in turn connected to another switch (C) which connects to end users. Graphically:

Serv----A----B----C----users

The point of debate comes in with regarding to where users can connect on this network. One viewpoint has it that they may only connect at point C, with additional switch/hubs branching off at points A and B (switch/hubs D,E,F, and G).
|-F |-G
Serv----A-----B-----C-----users
|__D |__E

The other viewpoint holds that users may also connect at point B.

|-users
|-F |-users
Serv----A-----B-----C-----users
|__D |__E

Who can speak authoritatively on this issue and explain the why's and why-nots (or even give an overview of 5-4-3)

Further, I noticed in another question, a response seeming to limit 100MB switch to Switch distance to 3 meters. Is this correct?

7 replies

by keith.abbott

Please note that the graphic representations got a little scrambled. F and D should be above and below A, G and E should be above and below B, and the additional users should appear on B as well

by ckilday

Keith,
I don't know about the 5-4-3 rule, but the situation you're describing sounds just like Cisco's preferred design methodology where you have three layers: core, distribution and access. Core would match your switch A, distribution matches switch B and access matches switch C. That works for routed networks, but I don't believe it's really applicable to switched networks. Our data center has two core Cisco 5000 switches with routing switch modules (RSM)in them. These switches connect to the server farm and other core network devices. They have mostly high bandwidth interfaces to handle the traffic and their focus is on providing fast access to those services. They connect to other Cisco 5500 series switches on individual floors of our two building complex. The individual users connect to those switches. Those B switches offload the processing of individual user requests from the core (A) switches so they can focus on their primary task. By keeping the users as close as possible to the core, we improve response time and access to services. That's why they access at the B layer. Your C layer switch will only add latency.
As for the distance limitation on 100Mb between switches, I think you're talking about some short copper jumper cable between stackable hubs/switches. We run 100Mb to the desktop over Cat5 copper cable at distances up to 100 Meters and we can run 100Mb over multi-mode fiber up to 900 meters or so. Hope this helps.
- ck

by tbradley

1st, let me answer the last question. Minimum cat 5 cabling distance is 3 feet, though this may be pushed a little. Maximum cable length for 10 and 100 mps is 100 meters.

2. The 5,4,3 rule really doesn't apply to switches. It was vitally important back when networks were designed around hubs, and even more vital in the days of 10Base-2 and 10Base-5.

3. Essentially, you can have 3 segments on your Lan, 4 repeaters, and 5 Nodes.

For this rule, a node was considered a device, a patch cable, and it's port on the
hub.

Ie

WS---Hub===Hub===Hub===Hub---Server

in this illustration, === equals segment,
Hubs=repeaters, and WS/Hub, Hub2 and Hub3, and Hub4/Server all equal nodes.

This rule has pretty much gone away since
switches have become so widespead, and
you can virtually expand your network
indefinitely.

A second Illus

WS--Hub===Hub===Hub===Hub---Server1
|
Hub===Hub===HUb===Hub---Server2

The workstation would not be able to
talk to Server2, nor could the 2 servers
talk due to the number of hops.

by tbradley

Getting the graphics down is hard.

In answer to your question, you can chain
users off any port you want, because you are
dealing with switches. In my Illustration, you can see that the max is 4 hops.

Hub1 to Hub2, Hub2 to Hub3, Hub3 to Hub4 and
Hub4 to the Server. You can place a user on
anyone of the hubs, but if 4 hops is exceeded, the user can't see the server.

Hope this clarifies it a little for ya.

This discussion is closed to new replies.