regarding leaf spine uplinks

[cowser1979]

Is there a specific design rule that dictates only 1x 40Gb uplink per spine switch?  I would think that for the purposes of growth, one could continue to add uplinks from existing leafs to existing spines, so long as the backplane of the spines and no. of front facing ports aren't exhausted?  Also, is there a specific design rule that demands the same number of uplinks per leafs?  The way we are building out our DC is per 9k leaf pair in a UCS domain (several chassis in a domain).  Would it be possible to have some leaf pairs with more spine uplinks to provide greater upstream bandwidth?  I realize that traffic going east to west to other pods with fewer uplinks may suffer, but I was curious if the APIC is capable of handling that?

thanks

[Tomas de Leon]

AFAIK, there is no design rule that dictates on 1 x 40gb uplink per spine switch.  As you mentioned, if you need additional bandwidth for east to west traffic, you add additional uplinks.  The concern or limitiations is port density and additional spines may be needed.  Symmetrical bandwidth topology of links between leafs and spines is optimal.  Asymmetrical could cause traffic issues if the links are fully saturated..

[cowser1979]

Thanks!