05-21-2020 12:50 PM - edited 05-21-2020 12:51 PM
Hello,
I created the attached diagram to illustrate what I am trying to accomplish.
When designing the uplink/downlink connections between a pair of C6509 switches configured in VSS domain and a pair of C9606 configured is SV (StackWise Virtual) domain, which option is preferable: option 1 or option 2?
Note: If you aren't familiar with the SV technology, imagine having two VSS domains instead.
Also,EIGRP is the dynamic routing used.
Option 1: Using L3/routed links criss-crossed between each member of the VSS and SV domains
Option 2: Using L3/routed links to directly connect each member of the VSS domain to the other member of the SV domain.
Thanks in advance,
Best, ~zK
05-21-2020 01:05 PM
Both are VSS / SV ( work same way and work as Single Switch, Same can be achieved in both the topions , but i chose Option 2, making L3 Port-channel terminativing the connection each switch.
05-21-2020 01:25 PM
Thanks for the input.
I plan on using ECMP (individual L3 connections) and not L3 Port-channels. If I chose option 2, will having individual L3 connections pose any issues?
Best, ~zK
05-21-2020 02:17 PM
what is the use case here for going each link as L3, if same can achived in better way.
VSS or SV have single control plane so, i do not see any advantage here going each link. until you have any specific reason here.
But if you looking each link as L3, then i suggest to have Option 1, so cover mesh here.
05-21-2020 06:29 PM
I reconsidered both options and I think option 1 is more preferred. See attached diagram.
From an upstream packet flow prospective:
With option 2; if there were a failure situation with the primary switch in the VSS or vice versa(VS), then you will lose BOTH uplinks from the primary switch in the SV to the failed primary switch in the VSS and in effect the traffic flow will have to cross the SVL toward the standby switch (in the SV) and then upstream to the standby switch, which will become the primary switch, in the VSS. There will be an extra hop along the way(crossing the SVL). This is not an optimal path.
However, with option 1; if there were a failure situation with the primary switch in the VSS or vice versa(SV), then you will lose one uplink from the primary switch and anther one from the standby switch in the SV and in effect the traffic flow will travel upstream towards the standby switch, which will become the primary switch, in the VSS from either the primary or the secondary switch in the SV without having to cross the SVL. Traffic will NOT have to cross the SVL link. This traffic choses an optimal path.
Regarding your question as to why individual L3 links is preferred over LAG (L3-MEC), there is a number of reasons. With ECMP; troubleshooting a link is easier; load-balancing doesn't rely on a hashing algorithm, depending on the option/mode of the port-channel you use, will force traffic to flow through the same link, those are some reasons.
Let me know what you think!
Best, ~zK
05-22-2020 12:11 AM - edited 05-22-2020 12:13 AM
Personally, This not L2, most of the Link you are looking L3, so Failure domain or Switch not have major impact here. other than re-routing the packet.
Only concern here, VSS (physically 2) Logically 1
For better high throughput, still i go with Port-cjhannel option.(my preference)
more information can be find here :
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide