|
netrop2
3 posts
 |
I have a network with 4 multilayer switch (MLS) creating the core, like the topology below: R1 ===== R4 Since they support VLAN interface, I ask: What are the advantages and disadvantages of using VLAN interfaces to create the OSPF adjacencies between them using a single mask (/29, for example) instead of using differents point-to-point (/30) between the devices? Addressing using a single mask: R1: 192.168.10.1/29 Addressing using multiple /30: R1 - R2: 192.168.10.0/30 |
|
mkomon
35 posts
 |
Single mask /29: Multiple /30: |
|
netrop2
3 posts
|
Thank you! But beside these considerations, are there any others points? |
|
bluepackets
25 posts
|
I'm not positive on that topology drawing, but from the way it appears, if you use SVIs there's a layer 2 loop. So in addition to the stated OSPF convergence impact, you also now have a spanning tree issue. Spanning tree will put one of those connections in blocking, which then creates the possibility that one of those switches will have to traverse up to two other devices to reach the DR (say if R3 is the DR and STP puts R2's connection to R3 in blocking). Though this will not necessarily affect anything in normal operation, it is certainly not a desirable situation. The question you have to ask yourself is: do you need layer 2 connectivity between these 4 devices? If you don't, then don't bother with SVI's. If you do, then you have two good options: use a different VLAN/SVI on each device to form OSPF adjacencies to the others and avoid the STP loop entirely or put in both a layer 2 and layer 3 link between each device. |
|
netrop2
3 posts
|
Bluepackets, thank you for your considerations. Yes, I'm gonna use different VLANs on each device, so I won't need STP in this situation. Actually, my point is about the addressing. Why would I use layer 2 conectivity with a single mask (/29, for example) between all devices instead of using point-to-point (/30) addressing between each device? Is there any advantage? Thank you again. |
|
bluepackets
25 posts
|
If the SVI's are different, thereby putting them in different VLANs, you've split a subnet between different VLANs, which is definitely not a best practice. As to why would you connect devices in that fashion and address them like that I have no idea; I can't think of any advantages. If you were to do that with more than one subnet and have routing adjacencies on the SVI's on both, I guess that can be argued. By that, I mean in this fashion (this is simply an example): R1 ==== R4 Each device is connected to each other device with a layer2 ether-bundle trunk of some quantity of links. Reserve two VLANs for routing, say 10 and 20. Pick which device you want as both the root for VLAN 10 as well as the DR, say R1. Configure the bridge ID for that VLAN on R1 to a value lower than default and the OSPF priority to a value greater than 1. Also, configure the OSPF priority to 0 on the other devices. Carry that VLAN only on the trunks between the DR and it's neighbors, or R1 to R2, R1 to R3, and R1 to R4. Pick a different device to use as both the root for VLAN 20, as well as the DR, say R2. Do the same as you did for R1 (configure bridge, OSPF priorities). Trunk whatever other VLANs you need between each. Now, you have a redundant layer 3 topology over a guaranteed loop free layer 2 topology for those VLANs. If the DR/root of one of the VLANs fails, the remaining DR/root will be used (assuming proper routing configuration). Obviously, a double failure of say R1 and R2 will wipe out layer 3 connectivity between R2 and R3, as an OSPF priority of 0 would prevent either of becoming the DR. You could go further and expand it to 4 VLANs and configure each core device as the DR/root of each, though. I would probably not even bother with even this unless you need to. |
Viewing 1 - 6 of 6
- 1