Advise on OSPF design

[rasmusan1]

Hello,

I am designing a network for low-bandwidth communication - resilience is in favor over performance.

Customer has their own fibres between sites.

Redundancy based on Layer 3 routing. this is a quite small network, with around 2 core switches, 2 distribution switches, 5 access switches with need for redundant connection to distribution switches, and ~30 access switches that does not have need for redundant links to distribution.

- Core based on Cat9500 (with stackwise virtual)

- Distribution based on cat9300,

- Access based on IE3300 (for switches with need for redundant links to dist = Layer3), and IE3100 for switches with no need for redundant uplinks.

Now I need advise for OSPF config - do you see any issues in defining this as one single Area 0 with this low number of switches? And do you have any other

Simpel topology as as shown below - all links as L3, unless those notes as L2 (no all access switches are shown in drawing)

Thanks in advance

/Rasmus

[paul driver]

Hello

---

@rasmusan1 wrote:

Access based on IE3300 (for sites with need for redundant links to dist = Layer3), and IE3100 for sites with no need for redundant uplinks.

---

You topology does not show both sites does it?,
Also can you confirm - L3 access-edge for both sites (with L3 resiliency at one site)  

[rasmusan1]

sorry for confusion - with sites I mean switches (corrected that in post)..... and yes L3 all the way to access switch, for those switches that need redundant uplink to distribution switches (thus IE3300 for those, as this supports OSPF).

The switches that don't need redundant links to distribution, those will be L2 switches (IE 3100) due to cost...

[paul driver]

Hello

---

@rasmusan1 wrote:

 and yes L3 all the way to access switch, for those switches that need redundant uplink to distribution switches (thus IE3300 for those, as this supports OSPF).

The switches that don't need redundant links to distribution, those will be L2 switches (IE 3100) due to cost...

---

So why not have L3 access-edge at both sites, even if you do not require link resiliency at one of them you would not need to worry about spanning-tree at any site either-  Just a thought?

[rasmusan1]

sure that would be best scenario - however there is a quite significant cost difference between IE3100 and IE3300, so thats the reason customer would only buy L3/OSPF capable switches where link resiliency is required... make sense?

But do you otherwise see any issues with this topology, where all L3 switches in same OSPF Area? any issues regarding routing loops to be aware of, or should OSPF take care of that just fine?

[paul driver]

Hello
Apologise \i didn't fact that i tbh, my mistake...
I would suggest the link between sites be in the BB area 0 and then either site in itd own area non BB area x

[rasmusan1]

could you maybe elaborate i little on why you would do that? why not everything in single area when topology this small?

/Rasmus

[paul driver]

Hello
TBH, my suggestion was based on a two SITE topology, However you have a single site with dual cores/distribution and partial L3 access-edge I dont not see any requirement to have multiple domains using so just using the BB area 0 would be applicable in this scenario

spanning tree would still be only applicable in this instance between the distribution and the L2 access edge switches 

[rasmusan1]

That is good to hear, as this was also what I thought

And I should not get an issue with routing loops from the L3 Access connecting to the 2 distribution switches - correct? And you thin my the outlined links in topology looks correct?

[paul driver · Accepted solution]

Hello
Unless you start applying redistribution/filtering etc then by default If you have all the cores/distribution at L3 plus certain L3 access witches with dual links then OPSF should cost out to provide a preferred SPF path between all nodes with no routing loops

[paul driver]

Hello

---

@rasmusan1 
 with sites I mean switches 

 Just realised you dont have 2 sites it just the one site with links correct ?

[rasmusan1]

yes core is on single site - distribution actually is different sites.... not sure how this matters - if you look at my topology, that show how the cabling is to be done.

Only thing that could be different, is link from dist to core, could be MC-LAG and thus single L3 link, rather than 2 individual L3 links, if there is any benefit in doing that?

[MHM Cisco World]

Ospf can run between only core and dist not need to run ospf in access' as I see you use L2 (trunk) between core and dist

MHM

[rasmusan1]

Yes sure, but as you see for the example "Access2" switch, I need link to both distribution switches for redundancy, and would like to base that on L3/OSPF

[MHM Cisco World]

For acc2 you can use hsrp in dist1 and dist2 and use hsrp group to load balance traffic.

https://www.cisco.com/c/en/us/support/docs/ip/hot-standby-router-protocol-hsrp/13781-7.html

MHM