cfan73
asked on
Spine Leaf (CLOS) design and campus LAN consolidation
I'm looking for some input regarding some recent spine/leaf design proposals I've come across. While a "traditional" S/L design would be to connect the enterprise campus core via L3 connections to the DC leaf nodes, what's been proposed are designs where all LAN distribution would also connect to the leaves, the S/L becoming a unified/collapsed DC and campus core.
Traditional
Proposed
I'm looking to vet potential concerns around this design (if any). If a customer is also looking to handle Layer 2 extension/overlay across the S/L fabric to support VM mobility (a la VXLAN), does this present a problem? Or, would we simply define the ports that connect to compute resources as VTEPs and NOT the links leading to the LAN switching?
Again, looking to see if others are running across (or even deploying) these collapsed designs, where the before/after net effect is the removal of the existing campus core switching layer (and thus, savings in hardware expense).
Thanks again
Traditional
Proposed
I'm looking to vet potential concerns around this design (if any). If a customer is also looking to handle Layer 2 extension/overlay across the S/L fabric to support VM mobility (a la VXLAN), does this present a problem? Or, would we simply define the ports that connect to compute resources as VTEPs and NOT the links leading to the LAN switching?
Again, looking to see if others are running across (or even deploying) these collapsed designs, where the before/after net effect is the removal of the existing campus core switching layer (and thus, savings in hardware expense).
Thanks again
Aaron Tomosky
IMO big core stacks are going away, or collapsing into leaves (call it a leaf, call it a collapsed core, whatever). With 20g LAGs coming from distribution stacks, if you have enough 10g ports on the Leaves and the E/W traffic paths make sense, there is no need for an additional core. However you really have to map out your specific setup, and account for your physical environment to know what makes sense.
ASKER
@Aaron Tomosky: Thanks for your feedback. We have sufficient port density to handle this, and might actually recommend to move the IDF connectivity up to the spine nodes instead to eliminate LAN traffic over the leaves.
That said, I'm still curious about potential conflicts/"gotchas" that could arise by putting LAN and DC traffic on the same fabric, or by deploying VXLAN to support the E/W L2 functionality while supporting LAN/user traffic.
So, leaving it open for further input.
That said, I'm still curious about potential conflicts/"gotchas" that could arise by putting LAN and DC traffic on the same fabric, or by deploying VXLAN to support the E/W L2 functionality while supporting LAN/user traffic.
So, leaving it open for further input.
This question needs an answer!
Become an EE member today
7 DAY FREE TRIALMembers can start a 7-Day Free trial then enjoy unlimited access to the platform.
View membership options
or
Learn why we charge membership fees
We get it - no one likes a content blocker. Take one extra minute and find out why we block content.