OSPF Area

[hs08]

Hello,

I have topology as below picture where i have 4 location connected over WAN. Every sites have dual WAN and use OSPF for routing protocol and divided into 4 OSPF area, also we define the OSPF as point to point in the config.

With this topology i want to ask :

• It's needed we have up to 4 area or can i use only single area (only area 0). What is advantages and disadvantages using muti and single area?
• If every router on each site have different WAN speed example 30Mbps and 70Mbps, how the traffic flow will utilize the bandwidth. Is WAN with 70Mbps will have more traffic or both WAN only can have 30Mbps to load balance so WAN with 70Mbps will under utilize?

[MHM Cisco World]

Core routers (both) must be in area 0 (backbone area)

Other branchs router can be in one are or two' 

I will check last point for suboptimal path with lab update you today 

MHM

[hs08]

Yes core and routers in same area for each sites.

[MHM Cisco World]

this my approach
the BW  (in my lab I use Cost) make one edge router always use since OSPF dont support ECMP so Core always select the low cost (high BW) link 
this can tune in ABR by area range but for now check my topolgy and write your notes

MHM 

[paul driver]

Hello @MHM Cisco World 
Due to the the way your design is (basically a clos) and depending on the core  you would be preferable to use ECMP,

[MHM Cisco World]

If I use ECMP I face what @hs08  mention below
check is statement 

• If every router on each site have different WAN speed example 30Mbps and 70Mbps, how the traffic flow will utilize the bandwidth. Is WAN with 70Mbps will have more traffic or both WAN only can have 30Mbps to load balance so WAN with 70Mbps will under utilize?

[paul driver]

Hello
In that case, Most preferable would be to amend the reference bandwidth on all the rtrs to accommodate the higher link costs and then possibly alternatives would include costing out the less the preferred uplink so not to have it being used (like you have shown) or maybe apply some policy based routing for certain traffic to use the less preferred lower cost bw interface or even look into some performance routing (oer) between each area so not to underutilised those lower bw links

[MHM Cisco World]

using area range as I mention before can make some load balance between two edge routes of each site
this command have keyword cost which use to make Core router prefer one path than other 
i.e. change cost per prefix not per Link 
MHM

[hs08]

Hello @MHM Cisco World 

Thanks for your lab. How we can change the cost per prefix?

[MHM Cisco World]

If you see my topology I make Edges routers as ABR in such I can use area range and with Cost we can load balance for each prefix .

MHM

[paul driver]

Hello

---

@hs08 wrote:

• It's needed we have up to 4 area or can i use only single area (only area 0). What is advantages and disadvantages using muti and single area?

---

OSPF rtrs utilise a Link state database (LSDB) to compile a view of the whole network, every rtr in the same area shares this LSDB and it has to be synchronised, So if you had a large single ospf area with lots of rtrs then that would be a lot of route information and route calculation done every time a change is recorded within the network and then propagated to every ospf rtr in that single area. 
 
Each non BackBone area needs to be directly connected to the BB area (area 0)  and this is done by the interconnection  between area border rtrs (ABRs), and its these ABRs that allow the communication between its own area and the rest of the network

The advantages of having these non backbone areas connecting to the BB area via those ABRs is the reduction of that route propagation and synchronisation throughout the ospf domain, As it is then limited to only the rtrs within their own area thus reducing the time it can take to update all the rtrs in the ospf domain and naturally the LSDB would be smaller as only summary route information about all others areas routes is required to allow for a complete view of the whole ospf domain.

Lastly these non BB areas can be also become stub areas which then can reduce the LSDB for rtrs in that area even smaller to an extent they would only require limited route information to reachability to the whole network 

For example based on your topology diagram , The areas 100,200,300 could in theory become total stub areas thus the rtrs will just need to know about routes internal to their own area and receive summary default route so to reach outside their own area, So drastically reducing their area LSDB , something not viable the rtrs were in a single large area 0.

---

@hs08 wrote:

• If every router on each site have different WAN speed example 30Mbps and 70Mbps, how the traffic flow will utilize the bandwidth. Is WAN with 70Mbps will have more traffic or both WAN only can have 30Mbps to load balance so WAN with 70Mbps will under utilize?

---

Each ospf interface will have a cost calculated based in its interface BW, so the higher the BW the more preferred it will be, However by default I believe this calculation (reference bandwidth) is (100,000,000 bps* BW )meaning a 100mb link= ospf cost 1 so unfortunately this means any interface higher then 100mb would also have a cost of 1, even a 1GB or  100GB link which isn't viable as you could end up routing via an overloaded 100mb link whilst the a much high bw link is left idle.

The options you have would either to manually change this opsf interface costing throughout your network so to allow a route path via interface(s) that need to be preferred or not OR change this default ospf calculation to reflect the correct link interfaces speed within your ospf domain

The latter can be done by increasing the ospf reference bandwidth on ALL rtrs within your ospf domain will result in a much more optimal routing path within your network without having to manually change ospf interface costing to reflect a much higher interface on a rtr that could presently being calculated equal to an interface with a much smaller BW .

[hs08]

in my topology we use ospf point to point, example we add some subnet on core area 100, then this subnet will be advertise to both router on area 100, then both router will update to router on area 0. Am i right? 

So what mean "reduction of that route propagation and synchronisation throughout the ospf" on my topology? Can you give an example?

[Andi Bintang]

From the perspective of LSAs, if area 100 advertises a subnet, what happens is that when it wants to inject it into area 0, it only advertises the summary LSA (LSA Type 3) to all other areas. This behavior is different if they are within the same area, where they will have the entire LSDB information of that area.

And also for reduction route, in case at the "special areas," it is possible to use a default route, which reduces the route lookup for all router groups within this "special" area. These special areas include stub, NSSA (Not So Stubby Area), totally stub, and totally NSSA.

[paul driver]

Hello

---

@paul driver wrote:
So what mean "reduction of that route propagation and synchronisation throughout the ospf" on my topology? Can you give an example?

---

The separation of the ospf domain with area 0 and other areas reduces the LSDB size of the rtrs in those areas, its like auto summarization on the area boundary, only summary information of a network from another area is advertised in/out of that area, the rtrs in those non bb areas do not need to know the full picture of the overall network to have full reachability

[Andi Bintang]

Hello @hs08 

1. For what is advantages of each multi and single area, take a look at the core concept of Area in OSPF

OSPF provides scalability for routing tables by dividing the topology segment into multiple OSPF areas within the routing domain. An OSPF area is a logical grouping of routers. All routers within the same OSPF area maintain identical copies of the LSDB. Its done if you remember/know about LSA 1 and LSA 2 (in multi access network/there are DR/BDR election)

As the number of network links and routers in the area increases, the OSPF area grows. Although using a single area simplifies the topology, there are considerations:

1. Full SPT Calculation will run when there is link flapping within an area.
2. In a single area, the size of the LSDB will increase and become unmanageable.
3. The growing LSDB for a single area consumes more memory and takes longer during the SPF calculation process.
4. With a single area, there is no summarization of routing information

Summary:

Single area OSPF provides simplicity in configuration, does not require summarization, and is easy for troubleshooting and changes. However, it has limitations such as lack of scalability, performance issues, and, as commonly observed in real production, if there is link flapping, the Full SPT Calculation will restart because the LSDB size is larger and consolidated into a single entity.

On the other hand, multi-area OSPF might introduce complexity in configuration and troubleshooting but offers benefits such as scalability, the ability to summarize routes, and each area having its own LSDB. This means that if there is interface flapping in one area, other areas will not be affected, as they only receive LSA Type 3 (summary LSA) from the ABR.

2. For OSPF Cost Calculation and how to calculate the best path using this formula

Cost = Reference Bandwidth / Interface Bandwidth

By default in cisco the Reference Bandwith OSPF is 100mb/100.000kbps

1. Cost for 30 Mbps Link:

Cost = 100.000 / 30.000
Cost = 3,33
Cost = 4

Since the OSPF cost must be an integer, this result is usually rounded up. So, the cost for a 30 Mbps link will be 4.

1. Cost for 70 Mbps Link

Cost = 100.000 / 70.000
Cost = 1,43...
Cost = 2

This result is also rounded up. So, the cost for a 70 Mbps link will be 2.

In OSPF, the path with the lowest cost is chosen as the best path. Therefore, the 70 Mbps link with a cost of 2 will be selected as the best path compared to the 30 Mbps link with a cost of 4.

Additional Consideration :

• Reference Bandwidth Adjustment : In modern networks, the reference bandwidth is often adjusted to account for links with speeds higher than 100 Mbps. For example, if there is a 1 Gbps (1000 Mbps) link, the default cost with a 100 Mbps reference bandwidth would be very low. Therefore, the reference bandwidth is often changed using the command 
• Multiple Paths: If two or more paths have the same cost, OSPF can perform load balancing between them.

Best Regards,
Andi Bintang