Why link distribution switches?

[graham robinson]

Hi

With regards to the Access/Distribution/Core layers of a network, most Cisco diagrams tend to show each access switch having links to two distribution switches, and each distribution switch having links to two core switches. The obvious point of this is redundancy, if any distribution switch or core switch [or the links between them] fail then traffic will be directed to another device via either STP/HSRP/OSPF/EIGRP, etc.

However, most diagrams also show an etherchannel between the two distribution switches. My question is why is this etherchannel necessary? Without it each distribution switch already has redundant links to the core so why add a loop? I realise STP will block this loop, but Cisco best practices seem to suggest removing reliance on STP as much as possible to improve convergence times, so wouldnt the network do just as well without a direct link between distribution switches?

Thanks in advance

[glen.grant]

  I think one of the answers is if your connection between the dist and core switch goes down or the core switch goes down  this will not cause hsrp or spanning tree churn as it will just take another route across that crosslink and out the other side .Routing  protocol convergenge is generally quite a bit quicker than waiting for hsrp, spanning tree etcc to settle out .  How does the diagram show it , as a routed  port channel ?

[graham robinson]

The diagrams in question are in the "Enterprise Campus 3.0 Architecture:Overview and Framework" document on the Cisco Design Zone:

  http://www.cisco.com/en/US/docs/solutions/Enterprise/Campus/campover.pdf

Figure 1 on page 6, figure 3 on page 10 and figure 4 on page 11 all show an etherchannel between the 2 distibution switches, along with redundant routes into the core. Figure 5 on page 13 in particular shows that the "best" design has redundant links to the core switches and an etherchannel to the other distribution switch [but doesnt specify if this etherchannel is layer 2 or layer 3].

Would a routed etherchannel be preferred to a layer 2 etherchannel?

[Vivek Ganapathi]

Hello Graham,

Though they haven't clearly mentioned whether it's a Layer 2 or Layer 3 port-channel, but they are talking about ECMP which clearly suggests the entire CORE-DISTRIBUTION is a Layer 3. So, Distribution-Distribution is again a Layer 3. HSRP traffic is flowing via the Layer 2 switches (Access switches) below to provide First hop redundancy to the hosts.

Regarding etherchannel between 2 distribution switches, i feel that's just for Layer 1 HA. Even we can have that between Core's. This is my personal feeling.

Thanks

Vivek