Hi,
Have a network design question. Basically we have 4 x X690's, two are in one DC and 2 are in another DC.
What the requirement is would be to LAG within the DC and across the DC's. A fabric solution would be perfect for this, but unfortunately this isn't an option.
One option would be to stack the X690's within the DC's and create an MLAG between the DC's, but I don't wont to using stacking in the core. Main reasons are that servers will be split within the DC so I wont to be able to upgrade the switches without resetting the whole stack, I would like to take advantage of both their control planes, and using MLAG makes use of all the switch CPU's evenly (fabric routing, slip VRRP).
Anyway, this is what I've come up with and would like to validate with the community or see if anyone has some better or other suggestions:
So here you have an ISC between each of the switches, and also OSPF P2P's between each switch. VRRP is split evenly between switches and fabric routing is being used.
The ISC 1, I add all the odd VLANs and all the switches connecting that side use those VLAN's, and visa versa for the other side.
The servers in the DC rooms reside locally only, and the VLANs added to the ISC between the DC switches only. DC VLANs needed across DC's I could use VXLAN.
I'm not sure this is an elegant solution, but on the surface it seems to provide an answer.
There might be a far better way of doing it, hence why I want to chuck it out there and see what you think.
Many thanks in advance.
Have a network design question. Basically we have 4 x X690's, two are in one DC and 2 are in another DC.
What the requirement is would be to LAG within the DC and across the DC's. A fabric solution would be perfect for this, but unfortunately this isn't an option.
One option would be to stack the X690's within the DC's and create an MLAG between the DC's, but I don't wont to using stacking in the core. Main reasons are that servers will be split within the DC so I wont to be able to upgrade the switches without resetting the whole stack, I would like to take advantage of both their control planes, and using MLAG makes use of all the switch CPU's evenly (fabric routing, slip VRRP).
Anyway, this is what I've come up with and would like to validate with the community or see if anyone has some better or other suggestions:
So here you have an ISC between each of the switches, and also OSPF P2P's between each switch. VRRP is split evenly between switches and fabric routing is being used.
The ISC 1, I add all the odd VLANs and all the switches connecting that side use those VLAN's, and visa versa for the other side.
The servers in the DC rooms reside locally only, and the VLANs added to the ISC between the DC switches only. DC VLANs needed across DC's I could use VXLAN.
I'm not sure this is an elegant solution, but on the surface it seems to provide an answer.
There might be a far better way of doing it, hence why I want to chuck it out there and see what you think.
Many thanks in advance.
