Hi,
I am considering to plan for the stretch fabric design over full meshed or separate fabrics in the three data centers located within 300 meters of range from each other and possibly huge fibers are available to interconnect them.
All three data centers will be highly populated and planning to have around 60 leafs per data center.
I have following concerns If I go to stretch fabric design:
1- Can we stretch ACI fabric in three or more different DCs within 1 KM?
2- Can we have pair (1+1) of APIC controllers in each site in the stretch fabric design and we have total 3 data centers? So cluster of total 6 APIC in 3 sites.
3- How many transit leafs can we have for the stretch fabrics on each data center?
4- What are the other limitations we can see in stretch fabrics?
5- Current limitation of number of spines are 6 in one fabric. How many spines in total will be supported in one stretch fabric?
Below are the benefits I am looking at when enable stretched fabric:
- Single point of management/configuration
- No legacy workaround between the data centers in case of workload mobility between the DCs.
- Any subnets anywhere in the whole fabric in all DCs.
- Controller cluster is split brain concept
- Stretch Vcenter/clusters between the DCs (Though It will be solved in vshpere 6.x)
- Supported 200 leafs per fabric + FEXs
Note: I can understand that Cisco will have super controller in future to manage multiple APIC controllers/fabrics.
Regards,
Anser
