I have three sites A, B, and C. A is connected to B using 5.7 backhaul. A is connected to C using same 5.7 backhaul. I want to link B to C using a 5.7 backhaul as well so that in case any one link AB or AC is down then the other link will ensure data flows uninterrupted. Will this work? if not what other alternatives exist? Greg
Although I do not totally understand what you have setup, I do know that you will need to use a 5.2 or 2.4 BH to link B to C.
My question is more about the IP traffic than RF.
Let me explain: I have a site (Site A) linked to two other sites (Site B) and (Site C)
My question is what will happen to the data flow if I link B to C, effectively having a ring/loop where data is transmitted simultaneous from A to B directly and A to B going through C.
I hope I made myself clearer this time
You do not want to create a loop in ethernet! :shock:
Just for fun, take a normal ethernet switch, connect a short cable from one port to another (loop), and connect it to your network. Wait a few seconds. The first multicast or broadcast packet that comes through will cause all ports to be locked up (well, almost). Unplug the switch from your network and watch - the multicast packet will travel around the loop forever, or until you break the loop or power the switch down. You don’t want that on your network.
Use switches that support Spanning Tree protocol. They will detect loops and down the redundent links until it’s needed. I am using D-Link DES-3526 switches and the Spanning Tree works great! If one link fails the switches automatically switch over and route data the other way.
I love technology! :lol:
Thanks for the info. I will get on it right away. Will keep you posted.
Some manufacturers also support proprietary protocols for redundant links. The primary benefit over STP is quicker recovery time from a failure on one link; seconds versus tens of seconds.
Another possibility is load sharing between links. This may or may not work, at Layer 2 anyway, when the separate links will likely have slightly different throughput and latency rates. If this does work, there should be no recovery time.
3Com has switches that support both of these features, but they cost 3-4 times the D-Link; maybe $1000 versus $300 for 24 ports.
That is true - no load balancing with STP. That said, you can force the links to one side or the other to use resources, but one link will always be unused at some point.
Regarding failover times, with advances in STP, Rapid STP has lowered the times it takes to find a down link so my network will switch over in a matter of 2-3 seconds. When that link comes back it will switch back in the same amount of time with no packet loss.
How would a multicast packet cause this to occur? I tried it out, but I didn’t use a seperate switch. I just took our existing office switch and ran a cable from Port 2 to Port 4. Within seconds the switch started going crazy and something started to beep. Don’t know if it was the switch or something else!
What exactly causes this to happen? What is stored in the MAC cache for these two ports? The MAC address of the switch? I would look for myself, but the managed switch it connected to the Canopy network!
Ha Ha Ha!! I should have said use a spare switch. Sorry. When that many packet go out of each port there isn’t much room for other traffic, knocking the network down.
Anyway, a multicast packet comes into a standard switch. The switch looks at and says “well, that should go ot every port”. So it sends it out every port. Then, it sees the packet again at the next port (in a loop) and the first port. The switch says “Well, that should go out every port”. So it sends it out every port. And so on and so forth. The perpetual machine has begun…
Anyway, a multicast packet comes into a standard switch. The switch looks at and says "well, that should go ot every port".
Ah, I see, for some reason I was thinking you said unicast in your previous post, but you really said multicast - that is what was confusing me.
Thanks for the explanation.
No problem. I actually learned that lesson the hard way.
:shock: “HEY! Where’d all this traffic come from all of a sudden?!” :lol:
It’s still neat to show people. I just grab a small 5 port switch and let it fly. It gets some strange comments from people that don’t understand.
In all my sites the BH plugs into the CMM. If I want to use the D-link or 3com switch how do I connect my BH units? Do I connect main BH + redundant BH to switch then connect switch to CMM? Will this affect sync settings in BHs?
The CMM’s Ethernet switch can be replaced without affecting sync. Sync is provided on a separate cable in the CMM2, and on different pins in the CMMmicro.
Be careful, however, about installing a different switch inside the CMM cabinet. Desktop and rackmount switches typically aren’t designed to operate reliably at extreme temperatures, hot or cold.
Thanks for the info and tip. Will get on it right away.
As above is right, you have to use Spanning Tree (SPT) capable switch (if you need faster redundancy, look for rapid spt capable switch). Another way (if you are having different ip subnets at sites) is OSPF. You have to have 3 routers, all with 3 legs (eth0 and eth1 connecting to BH, eth2 connecting to local clients.