01-19-2018 05:13 AM - edited 03-01-2019 05:25 AM
Just wanted to get some ideas or best practices on how to extend the ACI fabric. We have 2 spines and 6 leafs in the fabric in a rack in the data center. Need to expand the fabric to accommodate more devices but the rack is in the other side of the data center in a different row. Would you normally just run the fibre uplinks from the new leaf switches back to the rack with the existing spine switches? We are planning to add another 6 leaves so will need to run 12 fibres.
Or is there a more efficient way? is this what multi pod is?
01-19-2018 07:39 PM
Hi maleate,
Are you only planning to add leaf nodes to your existing fabric? If this is the case your only option would be to run cables from your six leaf nodes to your 2 spines.
Now if you were to add 2 other spine you could do stretch fabric connection (see link below)
Strech Fabric
Multipod is also another option but that would required to configured IPN (inter pe network) to interconnect the fabric pod as well as to have additional spines (see link below)
Multipod
01-23-2018 05:35 AM - edited 01-23-2018 05:37 AM
Thank you for the reply. I was not aware of stretched fabric earlier.
So am I correct in saying the only real benefit of multi-pod over stretched fabric is the separated forwarding control plane in multi-pod?
And within a single data center is there any real downside to extending the fabric with new leaf nodes and running the cabling back to the spines? This would be the simplest and potentially the cheapest option.
01-23-2018 08:39 AM
Regarding Ng your first question.
“ACI Stretched Fabric and its natural evolution named Multi-Pod.” “Those design options are usually fulfilling the requirements of interconnecting Data Centers deployed in Active/Active fashion, so to offer the freedom of deploying the various application components across separate Pods. The entire network hence runs as a single large fabric from an operational perspective; however, ACI Multi-Pod introduces specific enhancements to isolate as much as possible the failure domains between Pods, contributing to increase the overall design resiliency. This is achieved by running separate instances of fabric control planes (IS-IS, COOP, MP-BGP) across Pods.”
For your second question:
The only downside I can think of is if the location of your new leaf nodes are very far from the main fabric you would have to have the proper cabling infrastructure to interconnect these devices. In your case, since you are only installing your new leaf nodes a few racks away from the main fabric it would be the easiest and the cheapest solution.
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