number of ethernet hubs

[stu]

Hi guys
Quick question , does it mater how many Ethernet hubs you have on a plc system,I'm using slc plcs if that maters, thanks stu

 

[Mark Buskell]

For a switch what matters is how they are configured such as the mask, gateway, spanning tree and other variables.
For a hub, the more hubs you have the more broadcasts or traffic.
I have had 6 or so hubs on a site at one time but due to the traffic most of them where replaced with switches like an N-Tron or Hirschmann.

 

[Timbert]

What really matters isn't the number of hubs but the amount of traffic. If you have very little traffic having 20-30 hubs wouldn't be a problem. If you have lots of traffic having 2-3 hubs could be a problem.

Hubs broadcast traffic to everyone. Switches send data to who needs it. Using lots of hubs means to sending data to a lot of devices that don't need or want it.

If your only traffic is a 1 Hz keep alive signal, no worries. If you're running a fast update rate motion program you're gonna fill your data pipe with a lot of data that only one device needs.

Personally, I avoid using hubs at all. I just use switches to manage the data flow.

 

[Ken Roach]

Stu, are you thinking of the old 5-4-3 rule of Ethernet co-axial cable systems ? That applies only to co-ax cables, which have been replaced by twisted pair wire for decades.

If you are hinting at 'I am having a problem and think the number of hubs or switches might be the issue', tell us more about the system.

As said above, it's not the number of devices but rather the amount of traffic that matters when measuring or designing for network performance.

 

[Bit_Bucket_07]

I'm curious as to whether you're using the word "hubs" to describe "switches". As has already been pointed out, hubs broadcast every incoming packet to every port unconditionally, thereby creating redundant traffic, which can create a problem for any network. You'd be hard pressed to purchase an actual "hub" these days, as most electronics outlets will only stock "switches".

There is a limit to how many switches or hubs can exist between two Ethernet transceivers and it has to do with latency. Each data transfer through a switch (or hub) involves a small time delay. If a packet makes too many transitions prior to reaching it's destination, then it will time out and be dropped. A packet's latency time stamp is updated every time it passes through a switch (or hub).

 

[robertkjonesjr]

Are you sure this is only for coax? A true hub is a shared medium too so I would have included cat5 with hubs requiring the same rule. But why are we still using hubs? Are you sure they aren't switches? I'm not sure o can go into my local best buy and buy a hub if I wanted too.

 

[Ken Roach]

I'm not an Ethernet expert, but I think you're conflating the Time To Live (TTL) value with the 5-4-3 rule.

As I understand it, the TTL value is decremented by routers, not by switches or hubs. You never see it decremented when you are diagnosing a small LAN.

I will postulate that if you had an infinite set of hubs, connected with uplink ports, that a packet would continue to travel from hub to hub indefinitely. The signal is re-generated at each port, and there is no TTL decrement that would cause a hub to not send the packet on.

Before we spiral into a debate or discourse, I think we should take a step back and have Stu explain some more about his question.

 

[Bit_Bucket_07]

I'm not an Ethernet expert, but I think you're conflating the Time To Live (TTL) value with the 5-4-3 rule.

As I understand it, the TTL value is decremented by routers, not by switches or hubs. You never see it decremented when you are diagnosing a small LAN.

I will postulate that if you had an infinite set of hubs, connected with uplink ports, that a packet would continue to travel from hub to hub indefinitely. The signal is re-generated at each port, and there is no TTL decrement that would cause a hub to not send the packet on.

Before we spiral into a debate or discourse, I think we should take a step back and have Stu explain some more about his question.

I'm no Ethernet expert either, and I'm more than willing to concede that you are correct. At any rate, I've never personally encountered a scenario in which Ethernet traffic was halted due to an excess number of switches and/or hubs utilized in the network.

 

[robertkjonesjr]

It's not the infinite travel of the frame, it's the collision detection part of CSMA-CD. We haven't used hubs in years, and most others don't either so this issue really isn't visible anymore as it is an artifact from the early days. With the advent of cheap switches, it's really tough to actually find a hub. From time to time we do look for them as network taps and even if I find a new one that is labeled 'hub' I don't believe it until I try it. Some of the low-end SOHO devices are mis-labeled as hubs but are really switches, operating at layer 2.

 

[Ken Roach]

I'm going to go ahead and retract that postulation, as I was never really familiar with the '5-4-3' rule at a low level where a true 'repeater hub' was used.

Bit_Bucket is correct, and it's possible that Stu really does have a system using repeater hubs.

 

[robertkjonesjr]

A packet's latency time stamp is updated every time it passes through a switch (or hub).

Sorry to be stubborn, but where is this timestamp? Ken's approach for IP routing at layer 3 with the TTL is what I am familiar with for pruning traffic that has been around a while, but that is only through routers. Wireshark will show timestamps, but they are not inherent to the Ethernet frame, only to the Wireshark and system instance grabbing them and measuring them. TCP has an option to include timestamps in the header, but hubs and switches don't operate at that layer, so should never care (and it's somewhat unusual to see them in use anyway, in my experience).

 

[stu]

Sorry guys I ment switch , I have 5 plc's slc which all connects to a switch (SPIDER II 8TX Rail Switch) at one end of the line and I would like to add another switch at the other end of the line so I can adjust the program where I'm working , so didn't know what the rule are with 2 or more Ethernet switches on one network thanks guys