Nortel ERS 5520 PwR Switch

Hi Charles,

Let me try to answer our questions (which are very good by the way);

1) The end device (laptop or desktop) will not understand the 802.1q frame and will just drop the frames leaving any device you plug into the PC port on the phone unusable. In essence this command just tells the switch to leave the 802.1q headers on the voice VLAN traffic so the phone can identify those frames but strip the 802.1q headers for the PC traffic so the PC won’t freak out – it’s not expecting an 802.1q frame.

2) These are two very different protocols. MLT allows you to trunk two physical links into a single link at Layer 2 for additional bandwidth and additional redundancy should one link fail. VLACP is a method used to detect a communication problem over a link and mark the port as down so you don’t end up switch/bridging packets across a dead uplink – an uplink that has link but there’s nobody home on the far end. I use VLACP where I have Ethernet Switch 470s because the GBICs don’t support autonegotiation. Without autonegotiation there’s no ability to detect a far end failure – say a single fiber strand breaks, one switch will still have link while the other won’t have link. VLACP would detect that loss of connectivity and mark the port as down keeping your network from switching/bridging/routing traffic down a dead link where the packets would ultimately be lost forever.

3) I did include Spanning Tree commands. I recommend that everything use “fast start” because of the auto MDI/MDI-X feature where an end-user or confused technician could easily put a loop into your network by mistakenly cabling two ports together. Search for “spanning-tree” above, here’s the reference;
5520-48T-PWR (config-if)# spanning-tree learning fast
Note: you should NOT run Spanning Tree on your MLT ports!

4) Spanning Tree as a system wide protocol. We use to run Spanning Tree across an ATM LAN where it sometimes took 90 seconds for Spanning Tree to converge. We employ a few layers of protection; the first is at the closet switch which I discussed in the answer above using “spanning-tree learning fast” on all edge ports, the second we use SLPP (Simple Loop Protection Protocol) on our ERS 8600 cores and lastly we use CP limiting on the ERS 8600 cores which will shutdown an uplink if too many broadcast or multicast frames start flooding the network from that specific uplink.

Great questions and thanks for the comment!

You can run STP across a MLT but I would not recommend it. So my advice would be to disable STP on all ports which are going to belong to the MLT, on both switches.

Good Luck!

You’re keeping me busy there Michael! :)

The rate limiting feature is built into the ASCIC hardware so there’s no reliance on actual switch software – which is a good thing. While your basically correct, you’ll generally know soon enough that there’s a problem. While rate limiting will keep the majority of your switches reachable/manageable you still going to experience all sorts of MAC/FDB issues because of the loop. If you have a management system that is performing threshold monitoring that system will generally alert you to the surge in traffic. I’m currently using a combination of HP Open View and MRTG. In the majority of instances you’ll see SLPP kick-in and eventually CP-LIMIT will kick in at the core isolating the edge switch in question.

In a ultra secure environment you could configure MAC security (old school way) or you could go with a Network Access Control (NAC) solution which integrates with the latest Nortel switches.

With regard to your example… Spanning Tree enabled on the edge access ports will help save you 99% of the time in my experience.

Cheers!

Hello Samir,

I happy to try and help. What’s your question?

Cheers!

Hi Alex,

I’m not sure who you’ve been talking to at Nortel but you certainly don’t need STP enabled on a trunk just because you have multiple VLANs (802.1q) traversing that link. If you had multiple trunks between two switches without using a MLT/DMLT/SMLT configuration you would certainly need STP enabled between those switches in order to prevent the Layer 2 loops that would be present in such a configuration.

It’s my recommendation NOT to running STP between your edge and core switches. I definitely recommend you run it on your edge switches but not on the ports that uplink to your core (or distribution) network. You can run it if you chose to I just don’t find it very useful to-do so and there can be implementation differences between some vendors (example, Cisco floods BPDUs across all ports in an etherchannel configuration with Nortel only floods BPDUs across the lowest interface in a MultiLink trunk configuration).

In an SMLT configuration you CAN NOT run STP at all between your edge and core (or distribution) switches because it defeats the whole purpose of building a network architecture that is active/active as opposed to active/passive. In an SMLT design both uplinks from the edge are actively passing and receiving traffic, unlike when you use STP/RSTP/MSTP traffic can only traverse one of the uplinks while STP blocks the other uplink.

Hopefully that helps a little. Good Luck!