Migrating from EIGRP to OSPF

[dtran]

Hello everyone, hope everyone is doing well !!!

Thanks in advance for inputs/suggestions !!!! I am looking into migrating from EIGRP to OSPF inside my data center and needed some guidance.

I have been running EIGRP inside my data center forever and BGP on the WAN side. For the past few years I have the needs to exchange routes with 3rd party devices inside my data center and I've been able to BGP peer with these devices and do route redistribution with EIGRP and all has been working fine.

I am thinking to simplify routing inside my data center since these 3rd party devices support OSFP that way I only have to deal with 1 routing protocol inside the data center and not have to deal with route redistribution. I'll just have to learn OSPF which I have not worked with in the past, but that's not the concern (just another routing protocol). Is this the right path to take ???

I appreciate any inputs / suggestions !!! Please share any OSPF sample configs or documents that you may have. Thank you so much !!

Danny

[Joseph W. Doherty]

How "big" is your current EIGRP topology?  Do you do any thing "fancy" in EIGRP such as route summarization between routers or unequal cost multi-path routing, etc.?

If your current topology isn't "big" or "complex", likely you might be able to run OSPF using just a single area, which avoids much OSPF complexity (involved with multiple area usage).

Three things that come to mind, regarding moving from EIGRP to OSPF, first, if you use network statement within the OSPF process section, they work a bit differently from EIGRP's network statement.  For OSPF, the network statement is used to match interface IPs to be included into OSPF, not OSPF networks.  OSPF determines the network being used by the matched interface IP.  (Basically, matching interface IPs is much like matching an address in an ACL ACE.  You can even, if desired, match all the interface IPs using just a single network statement.)

Second, OSPF uses "cost" of interfaces to determine best path.  Cisco's OSPF implementations derive this cost based on the interface's bandwidth, using, by default, 100 Mbps for a cost of 1.  (On networks with links with interfaces "faster" than 100 Mbps, generally you want to change the default based on your interface with the highest bandwidth.)  Other vendors' OSPF implementations either require you to manually adjust/set each interface's OSPF cost, or they too will auto derive it based on interface bandwidth, but if they do the latter, very likely their base cost of 1 is for a higher bandwidth than 100 Mbps.  (Normally, you like all your OSPF routers to agree on the how they will cost interfaces, but technically, that's not required.  Can be confusing when it is not consistent and/or lead to path selection unintended.)

Third, other vendor's OSPF implementations, will almost always interoperate, but, for example, Cisco's OSPF implementation does things either not in the OSPF standard, or in the OSPF standard, but not really defined how they should be done.  Again, usually doesn't impact interoperation, but in rare situations, you can bump into unexpected issues.

BTW, believe @Richard Burts and @Peter Paluch are much more knowledgeable about EIGRP than I.  Perhaps one or both might join in for any other suggestions for a EIGRP to OSPF migration.

[dtran]

Hi Joseph, 

Thank you so much Joseph !!! very much appreciate your inputs !!

My EIGRP topology is small .. no more than a dozen devices at the moment. Plus I am replacing my core switch with a pair of Nexus 9Ks. I figure this is the best opportunity for me to make the change. But I just want to make sure I make the right choice in moving to OSPF. 

Thanks again Joseph !!! appreciate your valuable inputs !!!

Danny 

[Joseph W. Doherty]

Ah, with so few "routers", OSPF configuration should be relatively simple.  Likely, no more complex than what you've been doing with EIGRP.

BTW, on my web page for this thread, under "Related community topics", the first reference Migration from EIGRP to OSPF, @Richard Burts  has a nice posting mentioning how you can run EIGRP side-by-side with OSPF (and, by default, EIGRP has AD preference over OSPF).  Running side by side might be very helpful during a conversion.

[dtran]

Thanks Joseph !!!

A few benefits that I can think of for me to move to OSPF ..

1. Simplify my routing topology

2. Single routing protocol to manage across different vendors

3.  More scalable routing topology for future expansion

Anything other benefits Joseph ? Do you think that the right choice to do ? 

Thanks Joseph !!

Danny

[Richard Burts]

Danny

In your original post say say " that way I only have to deal with 1 routing protocol inside the data center and not have to deal with route redistribution." I believe that point is pretty convincing and supports the decision to transition from EIGRP to OSPF.

[Joseph W. Doherty]

Moving to OSPF because you need to interact with non-Cisco devices that don't support EIGRP, likely does make this a "right" choice.  (OSPF as a non-proprietary, multi-vendor, routing protocol, is probably the biggest reason why people move to it - as, in some ways EIGRP is superior to OSPF [though often not enough superior that most notice].)

Numbers 1 and 2 are likely true.  Number 3 is likely a "wash".  Both OSPF (especially Cisco's implementation) and EIGRP can scale to pretty large networks (like in the range of hundreds of routers).

Other benefits, besides inter-operability?  None that come to my mind.

BTW, biggest negative, likely just the learning curve involved with using a different routing protocol.

[MHM Cisco World]

All are nice for opsf except 

One big important point, 

The prefix filter

-Any router in eigrp can filter router, in opsf only abr and asbr have this capability, and to solve issue you need divide ospf into area

The prefix summary 

-Any router in eigrp can summary router, in opsf only abr and asbr have this capability, and to solve issue you need divide ospf into area

The huge database 

All router within one area have all prefix for all router, this utilize cpu and need high memory

That make opsf not good choice if you have already eigrp.

Instead treat each site as island and connect this island with bgp bridge, this solve many redistrubte issue between site.

That my opinion, 

Good luck freind 

[Joseph W. Doherty]

@MHM Cisco World raises a very valid difference between EIGRP and OSPF, but unless you're already have being doing individual EIGRP router route filtering and/or summarization, probably you don't need to use a multi-area OSPF topology.  (As you mention only having about a dozen devices, again, your topology likely is small enough you wouldn't need OSPF areas.)

BTW, a (very) old rule-of-thumb, is you can have up to about 50 routers per OSPF area.  On modern routers, OSPF, again especially using Cisco devices, seems it can go into the hundreds, for an OSPF area.  (Many variables actually impact how many routers that will work "safely" within an OSPF area.)

Also BTW, any Cisco OSPF router can filter what OSPF routes it places into its own route table, but such route filtering is not, and cannot, be applied to what routes a Cisco router sends to its neighbors (the latter can be done on ABRs/ASBRs).

[Richard Burts]

There have been some interesting points about whether/where you can do route filtering in OSPF as compared to the flexibility in EIGRP to filter routes on any device you choose, and where you can do route summarization in OSPF (only at area boundaries) as compared to the flexibility in EIGRP to summarize on any device you choose. I agree that both protocols can be implemented to scale up to support large networks (and that either of the protocols in a poorly designed network can have scalability issues). In what appears to be a fairly small and simple network described in the original post I do not believe that these differences would be significant. And I do believe that having a single protocol running in the data center that can support the third party devices is a significant advantage for OSPF.

[dtran]

Thank you so much everyone for your valuable inputs !!! 

The biggest challenge for me with this migration as Joseph has mentioned is the learning curve which I don't think is a big deal. I understand there is different area types in OSPF. Would I be using the backbone area 0 in my scenario ? 

Thanks you all !!!

Danny 

[Joseph W. Doherty]

For a single area OPSF topology, area number can be any number, i.e. you don't need to make it area zero.

However, still leaves the question, should your single area, now, be area zero or something else?

Well, assuming at some point you will need a multi-area OSPF topology, then you must have an area zero.  So, what I suggest, assuming you'll move to multiple areas, which would be easier to accomplish, converting part of your single area to area zero or to another area number?  (Usually, it seems to me, OSPF area zero is keep small, and if that will be the case for you, probably a bit less work to use something other than area zero now as there would be less configuration changes moving to a multi-area topology.)

[MHM Cisco World]

Use Area 0, In feature when you expand your network you dont care about the split area and split backbone area, meaning no need for virtual link.
where Backbone area 0 , if I need multi area ?
if you run 
Access-Aggr-Core 
between the Access and Aggr run any area other than area 0 
in Aggr-Core run area 0

[dtran]

Thank you very much everyone !!

I am running a pair of Nexus 9K as a collapse core. I think it makes sense for me to use area 0 ... this will enable for future expansion with less pain. Please let me know if you see differently or any issues/concerns. 

Thanks !!!

Danny

[MHM Cisco World]

are you run L3 
access-aggr-Core 
or 
Aggr is L2/L3 boundary SW ? If this is your case 
check the example in link below 
https://www.cisco.com/c/en/us/td/docs/solutions/Enterprise/Data_Center/DC_Infra2_5/DCI_SRND_2_5a_book/DCInfra_8a.html