Hi all,
Looking for some feedback here on why we’d need VXLan and the benefit of spine/leaf. We’re launching a complete DC refresh. Currently have 6 PODs each running a pair of Nexus 5Ks, all going back to a pair of 7Ks running fabricpath. All paths are active/active, and the fabric pretty much builds itself. Super easy to configure / maintain. We probably have 2000-2500 servers of which I estimate 70-80% to be virtualized. The number of physical changes in our DC is low.
Since fabricpath is pretty much dead, we need a new fabric technology that keeps spanning tree out of the backbone and allows us to utilize all links.
Today all our svi’s are on our 7Ks, we trunk all vlans to all the 5Ks and pretty much trunk all to all UCS FI’s as well. This gives compute the flexibility of vmotion without re-ip or networks involvement.
From my understanding with a spine/leaf infrastructure you wouldn’t connect anything other than leafs on the spine. You’d dedicate a pair of leafs for your border/wan connectivity, you’d dedicate a pair as your services leaf and make everything two hops away.
How is this better than terminating your wan Connectivity and services / firewall in the core making it only one hop for all your workloads?
How does vxlan make my DC better than how we have it setup today with basic vlan trunking?
We don’t have a requirement to vmotion offsite nor do active/active DR. No requirement to host the same ip space in a different site. We’re not multi-tenant and don’t need more than 4000 vlans.
Thanks for your thoughts!
No comments:
Post a Comment