cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
942
Views
0
Helpful
5
Replies

Layer2 topology technologies

colmgrier
Level 1
Level 1

Hi all,

UDLD

What type of UDLD will I apply for the below links. What is Cisco best practice for UDLD in a layer 2 topology?

a.Etherchannel(L2) ports between the Core switches.

b.Trunk dot1q uplink ports from edge switches to Core switches.

c.Trunk dot1q links between the edge switches(chained)

Storm Control/Broadcast suppression

I would like to apply storm control/broadcast suppression on all dot1q trunk ports to protect the core switches from broadcast storms on all VLANS. Is this best practice to apply this technology in a layer 2 topology? How should I apply this to the below links.

a. Trunk dot1q uplinks ports between edge switches and core switches.

b. Trunk dot1q links ports between the two edge switches (chained)

Edge switches (Chained)

What does Cisco recommend for the max amount of edge switches chained together using dot1q trunk uplinks. The network in question has more than 25 Catalyst switches chained together.

5 Replies 5

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello Colm,

I hope you are fine.

The drawback of chained access layer switches is that the throughput of both boxes is that of the uplink of the box nearest to distribution switch.

If they are two on each chain this can be acceptable if uplinks are at least GE or bundles of GE ports.

We still have in some campus a chain where the uplink is made of an etherchannel of two FE and the performance is poor with output drops on member links.

About UDLD:

the most useful advice is that UDLD timers can be too slow if you use any type of rapid STP RPVST or MST 802.1s even with the lowest timers settings it is slower then 802.1w Rapid STP.

If this is the case you should consider to use STP loop guard instead, that looking at STP BPDUs is not limited by timers.

We use STP loop guard combined with broadcast storm control 1.00 % on uplinks.

Towards router links you can think to use UDLD no STP messages should be received over them and the other device will not move its port to STP forwarding state either.

My only doubt is that probably routers don't support UDLD unless they are multilayer switches.

Then you can :

use UDLD aggressive to have the switch reacts to a detection by putting in errordisable the port.

We use it this way on campus still using PVST (no rapid).

About storm control:

we use it also on access ports combined with STP portfast and STP BPDU guard.

We use 1.00% on GE ports with no issue on /24 subnets.

If to be used on FE you could think to increase it to 5%.

Hope to help

Giuseppe

Hi Giuseppe,

Thanks for information contained in the above post.

Storm Control

Below is the storm control setting I have applied for the Core and Edge switch access ports and uplink trunk ports.

What is your opinion on the storm control setting for the access and trunk ports. Do I need storm control for unicast/multicast traffic.

note: The chains links consist of around 13 catalyst switches.

Core 6509

!- Access port

int g0/1

description 'Server Port"

spanning-tree portfast

spanning-tree bpduguard enable

storm-control broadcast level 1

storm-control multicast level 1

storm-control unicast level 1

Trunk port

int g0/0

description 'UPlink Trunk to edge 3560'

switchport

switchport mode trunk

swithport trunk encapsulation dot1q

storm-control broadcast level 1

storm-control multicast level 1

storm-control unicast level 1

Edge 3560

!- Aceess port

int fa0/1

description 'Pc Port"

switchport mode access

switchport access vlan

spanning-tree portfast

spanning-tree bpduguard enable

storm-control broadcast level 5

storm-control multicast level 5

storm-control unicast level 5

int g0/0

description 'UPlink Trunk to 6509 Core'

switchport

switchport mode trunk

swithport trunk encapsulation dot1q

storm-control broaddcast level 1

storm-control multicast level 1

storm-control unicast level 1

int g0/1

description 'UPlink Trunk to 3560 (Chained)

switchport

switchport mode trunk

swithport trunk encapsulation dot1q

storm-control broaddcast level 1

storm-control multicast level 1

storm-control unicast level 1

---------------------------------------------------

Chained switches

The topolpgy for this network consits of several chains of 3650 switches. Each chain consits of around 13 Catalyst 3560 switches.

Each link is 1GB from the Core to the Edge switch. What improvement can I propose to customer. (port-channel uplink, etc.....

What is the max number of access switches can you chain?

Customer Topology

6509-A link a (13 chained switches)

| link b (13 chained switches)

| link c (11 chained switches)

| link d (9 chained switches)

| link e (13 chained switches)

| link f (13 chained switches)

|

|

|

6509-B link a (13 chained switches)

link b (13 chained switches)

link c (11 chained switches)

link d (9 chained switches)

link e (13 chained switches)

link f (13 chained switches)

Hello Colm,

you have up to 13 switches in daisy chain with a single GE uplink.

I would propose a new single fiber based cabling and/or I would think of some forms of stacking if the C3560 support the stackwise technology and stackwise cable. For C3750 the stackwise cable is 32Gbps capable.

In the first case you may need to add 24 ports fiber based linecards to your core switches.

Or are the switches distributed in a long run and you cannot connect them to the core/distribution switches ?

Possible suggestions in the short term:

use etherchannel of two GE in all the chain if possible: if you have 24 FE users in each switch you have 300 users in a chain this currently makes a 1:3 ratio between available BW and wire speed on user ports.

This is your 3560 have 24 ports or 24 users. If they have 48 FE ports and 40 users each you are currently in a ratio 1:5.

Hope to help

Giuseppe

". . . if the C3560 support the stackwise technology and stackwise cable."

NB: 3560 doesn't.

Another possible option, to avoid the daisy chains, would be in place a distribution stack of 3750s, using the 3750G-12S model. In this stack, run fiber uplink from each edge 3560 to 3750G-12S device.

From the new 3750 stack, you can run one or more fiber gig connections to 6500s, or place a 3750 with copper in stack and you can use copper gig uplinks between 3750 stack and 6500s, or place 3750-E in stack and you can run 10 gig uplinks to 6500s. (BTW: such a "star" configuration will increase bandwidth to most edge switches and avoid daisy chain member failure blocking access to other downstream members [unless you've looped topology and are using STP].)

If you want to improve redundancy, run at least two fiber uplinks from each 3560 to different 3750 devices in stack in Etherchannel configuration (also adds available bandwidth to edge).

If you run routing on the 3750 stack, you can easily run uplinks from the stack to both 6500s and use both links activately.

Problem with the daisy chain is that all the switches are in seperate wireing closets.

Therefore I cannot change the network topology but would still like to apply storm control and udld on all figer uplinks.

Review Cisco Networking products for a $25 gift card