Apologies, another long one...
Wanted to talk about these two different design approaches for essentially the same thing. This neatly ties into a couple of my recent posts...
The first is using MLAG - which is the approach I've adopted in another network:
- The design entails four cores, two in different locations using MLAG (two locations shown by dotted red line)
- All four cores have a common VRRP VRID 100 and common VLANs shared between them
- Each pair of cores have their own VRRP VRID 10 & 20 for only VLANs in those areas
- Fabric routing mode is enabled on all VLANs
- OSPF is enabled on all VLANs as passive. Except /30 between each MLAG pair i.e cores in VRID 10 have a /30 between them as does the cores in VRID 20.
- OSPF is configured for broadcast and using a /29 address running between and joining each of the cores together at layer 3, although they run on top of a single lag.
The second option is using stacking. The design is a lot simpler, and that appeal alone might a draw many using it, but I have kept to the habit of keeping stacking out of any large core / data centre.
As you can see it offers exactly the same as above but just tilted around. You can do inline upgrades in both examples as each link is diversely connected.
So these are my arguments for each, but I'm interest in all pro's and con's the community could come up with both
MLAG Approach - although more complex....
- It offers the best (by a fair margin) the most optimal approach
- Traffic is more evenly distrbuted to each of the switches
- Each switch is independently able for forward / route traffic with out passing to a master switch
- I can better control traffic distribution across links via lag hashing algorithms
- I can simply bundle more links into the lag to increase bandwidth
- Can offer a more economical approach by using more cost effective links and simply adding more as and when required
- The approach offers more stability over stacking failures
- This approach seems more logical for switches that are more geographically separate i.e across the red line. Although possible with V-Stacking its probably not a good idea to have stacks joined so fair apart.
- Configuration is a lot simpler and easier to manage
- Possibly easier to add more switches to core by adding more to the stack, but this is perhaps not a good idea. MLAG design you can just lag to another switch
- Possibly OK approach for smaller sites.
- Possibly OK where all the cores are close to each other
Many thanks in advance