Can we get ride of STP by using Stacking and Dristributed EC ?

[jclerico]

Hello all,

we have 2x CBS3010 (ie C3750 like for IBM BladeCenter) named SW1 and SW2.

Both are connected with 1x uplink each to a central Catalyst 6500.

The topology appears like a triangle and STP is needed to break loops

.

Question1:

- if I stack both SW1/SW2 and create a Distributed EtherChannel (using both uplinks) towards the C6K..

- Do I still need STP ?

.

Question2:

- let's say that there is one BladeServer connected to SW1, and another one connected to SW2.

- Traffic received by SW1 (from downstream server1) is switched out via uplink1 to C6K

- Traffic received by SW2 (from downstream server2) is switched out via uplink2 to C6K

- NOW if I stack both SW1/SW2 + bundle both uplinks, how traffic is switch out to C6K in case of one port-channel member failure ?

.

Many thanks for your help

PS: STP is a problem for me because C6K runs MST with allx 4094 spread in 2 instances... and my low end switch cannot handle it

Rgds Jean-Charles Clerico R/S CCIE

[Reza Sharifi · Accepted solution]

Hi Jean,

Question1:

- if I stack both SW1/SW2 and create a Distributed EtherChannel (using both uplinks) towards the C6K..

- Do I still need STP ?

No, you don't, because when you stack the 3750s in the IBM chassis, they logically become one switch.  So, no need for STP.  The switches in the IBM chassis are usually 3110s or 3120s and as you already know they are stack able just like the 3750s.

- let's say that there is one BladeServer connected to SW1, and another one connected to SW2.

- Traffic received by SW1 (from downstream server1) is switched out via uplink1 to C6K

- Traffic received by SW2 (from downstream server2) is switched out via uplink2 to C6K

-  Yes, the traffic will be forwarded regardless of what blade is connected to what switch

NOW if I stack both SW1/SW2 + bundle both uplinks, how traffic is   switch out to C6K in case of one port-channel member failure

If one line in the Etherchannel fails the other link continue forwarding traffic.  There would be no down time at all.

If I remember correctly the 3100 series use source/destination Mac for forwarding packets. So, let say you have 2 blades in yours chassis, when blade one sends a packet it will use first link and when blade 2 sends a packet it will use the second link.  If you loose one link the other link forwards packets for both blades.

HTH

[jclerico]

Many thanks Reza

.

I am trying to figureout the different options to get ride of STP between the:

- Distribution Block made of C6500 or Nexus 7000

- Access Block made of 2x embedded Blade Switch Cisco

.

I am suggesting many design options which are:

1- If upper layer is virtualized (Nexus vPC peers and/or C6500 in VSS mode) than we could leave both Blade Switch Cisco unstacked and connected each of them with 2 uplinks ending on the above 2 chassis (LACP).

2- If upper layer is not virtualized (Nexus no vPC peers, and/or C6500 in standalone) than we could stacked both Blade Switch and connect this stack with 2 uplinks (LACP) to a single upper chassis

.

In option1: per vPC and/or VSS recommendations, STP is still recommended ...

in option2: I didn't find any Cisco recommendations regarding STP and think like you that it is still not needed here

.

Do you have some url and/or white paper saying that STP is not needed when stacking is implemented

Rgds JCharles