Hello
Apologies if this is answered already -
Currently I run a small (10 node) ACI fabric will all of its EPGs mapped to a stand alone BD, set to legacy flood and learn mode. There are no IP interfaces on any epgs,as all routing is done by other devices outside of the fabric (non integrated FWs and LBs and routers). I also do NOT utilize any contracts (since there are no L3 interfaces). Are there any documents or design guides somewhere that may shed light on scalability limits in this mode, vs the optimized forwarding model, where MACs/IPs are stored in the Coop DB on the spines? Or any other operation gotchas?
I would like to try to get an understanding, regarding any limitations (or dangers) that I am placing on my fabric, by staying in this mode of operation. I understand the basic differences between the two modes, but I want to make sure there isn't some unforseen limitation/gotcha, that I am missing. The only limitation that I could see so far, is that since I;m not taking advantage of the SPINE DB, I will be limited to the amount of TCAM in each leaf (9396Px)?
thanks
