Maximum Chained Switchs IEEE ???

[JeffG1]

In a situation where you have chain one switch off of another one and so one... what is the official IEEE standard for the maximum number of switches. I thought it was like 3 or 4... any one know the answer. This would be for cisco 2924 and 2950s

[konigl]

As long as you are not using Spanning Tree, in theory you could daisy-chain switches indefinitely. Although you would run into general performance problems with large, flat broadcast domains and broadcast traffic; and specific performance problems with systems on the outer fringes of this network trying to fight their way through all the uplinks to get to resources in the middle or at the opposite end of the network.

With traditional Spanning Tree enabled, the default timers assume a network diameter of 7 bridges or switches between any two computers. You can adjust the timers so that the diameter can be expanded, at the expense of network reconvergence time.

I think in another post I did elsewhere in the NetProf forums, I actually calculated the maximum diameter of an STP-enabled network, using the maximums on the timer parameters according to what the switches would let you tune them to. I don't remember what that number was, though. I'll post a link to it here when I find it.

Shared 10-megabit Ethernet had a 5-4-3-2-1 rule (I think it was "no more than 5 cable segments between any two computers, connected to no more than 4 repeaters, with no more than 3 of those cable segments having multiple --that is, more than two-- devices attached to it, with 2 of those segments being dedicated to inter-repeater links, forming 1 collision domain").

Shared 100-megabit Fast Ethernet was much more restrictive about the size of its collision domains, depending on whether you had Class 1 repeaters (in which case it was just one repeater/hub/concentrator) or Class 2 (where you could have two, but the interconnect cable between them was really, really short).

Bridges and switches were what you had to put in to connect multiple shared-media collision domains, and got you around these limitations. And this evolved into the switched-only LANs that most people use today.

Hope this helps.

[konigl]

UPDATE: Maximum STP diameter

As I mentioned in my previous post, there is an upper limit to STP network diameter based on how you tune the timers. But I didn't actually post the answers. So here they are.

In a post I did on August 12, 2003 (search for "konigl stp diameter", or follow this link: http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&CommCmd=MB%3Fcmd%3Dpass_through%26location%3Doutline%40%5E1%40%40.ee9e70c/5#selected_message) I mentioned how you could calculate what this is.

Using the STP Max Age (MA) maximum of 200 and the STP Forward Delay (FD) maximum of 200, and varying the STP Hello Time (HT) from the minimum of 1 to the maximum of 10 seconds, you can get a maximum theoretical STP network diameter of 97 (HT=1) or 79 (HT=10). But that incurs an awfully long reconvergence time whenever the Spanning Tree is disrupted.

Having said all that, it's really just good networking design practice to keep the network as flat as possible, minimizing the number of switch hops or uplinks that must be traversed to get to server-type resources, and balancing the network load as best you can to give everyone an as equal a level of access as possible.

Hope this helps.