Discussion Locked

100BaseT Maximum Hubs???

0 helpful
11 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.
Our network runs Cat5 UTP and has in total 6 D-Link 24 port hubs. 4 hubs are located in the same room as the NT server. When I try to link the fourth hub collision lights then appear on all the hubs. When only three hubs are linked there are no collisions and using PING returns four replies. When the fourth is connected PING only returns 2 replies and 2 timed outs. Is a 100BaseT architecture under the 5-4-3 rule as with 10Base2 or 10Base5? Our hubs are configured with uplink buttons and we are using straight cables to link them. Also the fourth hub does not have the uplink button pressed and a cable in that port. Should the uplink start from the fourth hub and goto the first or the other way around? Any suggestions would be appreciated.

11 replies

by jimbb

Ethernet is ethernet, no matter at what speed. So yes, the same 5-4-3 rule applies.

Silly question: can you dedicate some kind of a "master" hub, and fan out from there, so that each other hub is connected to the "master"? Of course it depends on physical locations and distances and so, but this way you could have as much hubs as your "master" hub has ports, without having to worry about interconnecting.

One more thing: on a 100 Mbps ethernet network, your maximum distances within a "collision domain" (=all nodes and hubs that are not separated by a router or a switch) is much more limited than in 10BaseT. Your new hub may just push you over that limit, and your collisions and poor connections are an indication of that.

Jim

by vh_s

I thought D-Link provide a daisy chain cable by the purchase of a hub.

To my understanding (remembering just vaguely), you could daisy-chain the hub with a max of 4 hub stacked altogether. These 4 hubs will act as one single-hub, therefore extending the 5-4-3 rule as these four hubs is thought as 1 object.

Why ? I think the daisy-chain cable provide a more bigger bandwith than the uplink cable, and therefore decrease the chance of collision.

Remember also that most 100BaseT is set-up to use half-duplex instead of a full-duplex connection.

by wanabedrsean

it depends on how you are cascading them the master hub theory would be more correct than cascadeing one after anoter but the daisy chain is the best soloution it gives all the hubs the same 100 mb fabric the majic word here is propagation delay thettl (time to live) isnt long enough for three hops through hubs if you are pinging and want to test this use the -i and -w switch with ping and see what results you get the -w is maximum reply time in milliseconds 1000 ms=1 second and see how long it takes to get a full reply a traceroute will also give you some helpful info. good luck

by mckaytech

Something is not right with that fourth hub - it sounds to me like the interface is failing and causing a condition similar to a "jabbering" NIC.

That having been said, I think one of the things I would do is to buy a 4 or 5 port switch and plug each of the hubs into that rather than daisy-chaining the hubs. In addition to having a possible speed benefit, it would also cut down the size of the collision domains.

paul

by jaymal1

Etherent is NOT ethernet!

The 5-4-3 rule ABSOLUTELY DOES NOT......let me repeat this for emphasis.....THE 5-4-3 RULE DOES NOT APPLY to Fast Ethernet.

Those who clain it does should refer back to the IEEE specifications.

The only commonalitybetween 10 and 100 is the Cat5 wiring spec. You can have ONE repeater in Fast Ethernet (It's either class 1 or 2). Most hubs that "gang" together for any ethernet do it through a backbone connection, not a repeating port.

You need to look at the manufacturer's specifications for the hub to see what they support. Remember to look for Fast Ethernet specs, not traditional ethernet specs.

There are bit time calculations you can perform to see if you are exceeding the round trip latency for the circuit (which is what it sounds like).

This discussion is closed to new replies.