How would you design a DC network with 4-8 switches per DC (total 2 DCs)? Seems like wasting money to build a complete spine-leaf architecture with so few switches. (We're mainly running hyperconverged stuff so we're not needing that many racks any more)
I'm thinking about doing basically a "ring" of the DC switches (theyre in MLAG pairs) and then connecting the other DC from one/two pairs depending how many fibers we can get. OSPF or BGP in the underlay and EVPN overlay.
Spine-leaf architectures are all about equal distance between every servers but do you see this kind of setup problematic?
No comments:
Post a Comment