1

When I search for OSPF real world application or implementation, I always get the explanation of how OSPF works or how to calculate the cost. My question is, is there any real life scenarios of how OSPF works?

2
  • OSPF is an open industry standard, while EIGRP is a Cisco proprietary protocol. – Ron Maupin Dec 18 '20 at 13:52
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer. – Ron Maupin Dec 31 '20 at 19:40
2

is there any condition where OSPF is better used than anything else (e.g EIGRP), or what kind of network is suitable using OSPF?

In a heterogeneous network, all the router vendors will support the open-standard OSPF. For EIGRP, you would need a Cisco-only network, and it would close you from using routers from any other vendors. It would also create big problems in the event you merge with another company.

RIP is also an open standard that is supported by most vendors, but it has severe limitations, such as hop count and convergence time.

IS-IS is another open standard routing protocol, and it found a niche in telecommunications companies.

Overall, OSPF is open and well understood by people in the industry, and it is trusted to do the job over a wide range of equipment and network sizes. Most shops want to use products that do not lock them into particular vendors, and their employees and prospective employees understand.

2
  • I think EIGRP became an open standard in 2016 very much like what YouTube did with QUIC. Correct me if I am wrong. – Bharat C Penumutchu Dec 20 '20 at 8:42
  • In 2016 (many years after the fact), there was an Informational RFC published describing Cisco's EIGRP. That was far too late to the party for it to actually find support among router vendors. EIGRP is essentially a Cisco-proprietary protocol because there is no industry-wide support for it. – Ron Maupin Dec 20 '20 at 13:34
2

Network topology might affect the choice of a routing protocol and make one protocol more desirable than another. The most common example is a large hub and spoke topology. In that topology, EIGRP can handle routing updates more efficiently than OSPF.

The bigger difference in protocols comes down to the practical implementations of them. As others have pointed out, EIGRP is a Cisco-proprietary protocol*, so that may be important if you have a multi-vendor environment.

Another example of practical implementations is OSPF vs IS-IS. Most large ISPs use IS-IS for their internal routing, even though IS-IS and OSPF use the same algorithms and have similar performance. From an operational point of view, IS-IS has fewer features, and therefore is "simpler" than OSPF. To many ISPs, that simplicity also means reliability -- there are fewer things to go wrong, which makes it more desirable than OSPF.

*EIGRP is a defined standard (RFC 7868), but hardly anyone outside of Cisco has implemented it.

0

OSPF is likely the most popular routing protocol within private networks. OSPF peers exchange information about the networks they are connected to, plus injected static routes. That enables fast and reliable route propagation.

3
  • thank you for responding! my question is more like is there any condition where OSPF is better used than anything else (e.g EIGRP), or what kind of network is suitable using OSPF? – ella Dec 18 '20 at 9:24
  • OSPF is preferred for private (trusted) route exchange because it's less complicated than the alternatives. OSPF puts little emphasis on security, so it's less suited for exterior gateway exchange. – Zac67 Dec 18 '20 at 9:31
  • The argument for fast convergence is not really cutting it for OSPF because it's supported with all protocols. Even late versions of RIP support triggered updates. For small-enough networks, RIP can be most efficient. – caveman Dec 20 '20 at 5:12
-1

1. OSPF

1.1 Marketing view

OSPF's advantage is made of multiple points that make it appealing:

  • More efficient than RIP with large networks thanks to OSPF being a link-state protocol, and due to the problem with RIP that it's too naive (it periodically sends the full routing table, which is very inefficient). Note that RIP's limitation is not due to fundamental flaw with distance-vectors routing protocols, but rather due to specific choices made within RIP. A distance-vector protocol can still be efficient, but RIP is just not one.
  • That it's an open protocol that's wide spread on routers from various vendors. This can allow you to incorporate various routers from different vendors with each other. But this is not enough by itself, since IS-IS is also an open protocol like OSFP.
  • Its configuration is based on IP addresses. This makes it easier for some people like IS-IS which is based on CLNS addresses for its configuration.
  • OSPF is more wide spread than IS-IS. In part because of the fact that IP addresses won against CLNS addresses. IS-IS was made by ISO for CLNS addresses and CLNS lost the battle. This battle loss is only historic, as IS-IS supports routing for IP networks, and ironically, IS-IS supported IPv6 before OSPF.

Due to those reasons, people go with OSPF. In other words, OSPF fills a point where it's more efficient than RIP for large networks for being a little harder to use (arguable) and being the most wide-spread open routing protocol after RIP.

1.2 Technical view

But is OSPF technically any better? The answer is no.

Both EIGRP and IS-IS are technically better than OSPF. OSPF has this needless requirement that all areas must connect to area 0 (backbone). IS-IS does not have such requirement. OSPF's requirement of having all areas connect to a backbone is going to limit the way your network growth in the future, and imposes a burden on routers inside your backbone area by requiring to make them computationally more capable (since all other areas talk to each other via the backbone). This is really a bad design by OSPF.

2. IS-IS

IS-IS on the other hand, does not have the backbone concept. You can attach your areas to each other without needing to have them go through a given area. If you want to have a backbone area, then you can arrange your IS-IS areas such that they end up all going through an area of your choice. But if you don't want, then you don't! IS-IS also solves the shortest-path-first problem within an area like OSPF. As said, IS-IS is also an open protocol, but it's configurations requires selection of CLNS addresses (not IP) which confuses some people, and IS-IS is less wide spread. IS-IS is more popular with service provide where the network engineers are more knowledgeable and can purchase higher-end equipment that implement IS-IS.

3. EIGRP

But, technically, the best in terms of computational needs is really EIGRP in my view. Unlike OSPF and IS-IS (which solve the shortest-path problem for all links in a given area), EIGRP distributes such computation across all routers inside a given administrative region, such that each router does its own part and passes his work on. Eventually EIGRP routers converge to the correct solution, without needing every router to solve the shortest-path-first problem for all networks in the administrative region (or area as named in IS-IS/OSPF). Such administrative regions are also defined more flexibly by EIGRP by means of route summarisation between routers (practically you can reach an identical effect to areas).

But the major drawback of EIGRP is the fact that it's not an open protocol. This is a huge problem as it will lock you, as a company, from freely replacing your routers by non-Cisco routers. Being locked to a specific vendor is a danger. This is why, even if your routers are 100% Cisco routers, I'd suggest that you avoid EIGRP in order to give yourself room should you one day stumble upon a moment where plugging a non-Cisco router is needed.

EIGRP is a purely distance vector routing protocol and is an example that RIP's problem is not due to a fundamental flaw in open-distance-ness, but rather due to RIP being too simple to handle large networks.

Extra nuggets

  • RIP can still be the most efficient protocol for stable-enough small-enough networks. If the routing table is small, then RIP's messages can be smaller than the link-state messages of OSPF/IS-IS (some of which use padding which makes their messages possibly larger than RIP's updates if RIP's routing table is small enough).
  • EIGRP is often told by Cisco to be a "hybrid" routing protocol, as a marketing trick to avoid the bad reputation of some distance vector protocols such as RIP. But technically EIGRP is a distance-vector routing as it purely learns by rumour.
  • The real hybrid routing protocols are actually OSPF and IS-IS, because they are link-state only within an area, and operate purely in a distance-vector manner between areas. In fact the concept of areas are made with them only because solving the shortest-path-first problem for too-large networks is too expensive. Hence the shortest-path-first is solved within an area, and the inter-area becomes the cheap distance vector method of learning by rumour from neighbouring routers of other areas. Worth nothing that OSPF and IS-IS being "hybrid" does not give them any special points. I'd argue that this hybrid-ness of OSPF/IS-IS is a downside as it implies increased complexity of protocol's implementation.

Summary

If I to rank routing protocols from purely a technical perspective, I'd say that RIP is needed since it's the most efficient choice for small-enough networks out of all of them. IS-IS is needed for being the best available interior routing protocol that is an open protocol. EIGRP is needed as the best interior routing protocol for those that swore the oath to never leave Cisco. But OSPF is just not needed from a purely technical perspective. OSPF is only needed due to the human factor (people being used to IP addressing in configs) and due to historical events that made OSPF become more wide-spread than IS-IS.

Note to new OP

Oh, and welcome to the world of computer networking! You will see confusing names due to historical reasons, more sub-optimal solution that exist today simply because people are just used to them, and people that blindly defend protocols as-if protocols are always right (which is not true by the way, protocols have mistakes). Plus some new cool tech that's being cooked in the labs.

That said, I have to say that you are doing the right thing. Your approach is logical and are asking the right questions. The world of networking has some accumulated mess due to historical reasons, and also has some real genius behind some of its protocols. This is why it is sometimes confusing as you ask yourself: why do we have X? Is X a genius that I am not foreseeing yet? Or is X an ancient mistake that we are stuck with? Or is it just some dudes that thought that everything must be in XML?

Buckle-up for the ride is gonna be fun!

4
  • Ones who down-vote this post: kindly specify your reason. – caveman Dec 20 '20 at 9:18
  • 1
    I have no idea who voted down on your answer, but remember that voting is meant to be anonymous, so trying to force a voter to reveal identity by commenting defeats the anonymity. I would imagine it is because the voter does not believe your answer, while very complete, addresses the OP's comment "my question is more like is there any condition where OSPF is better used than anything else (e.g EIGRP), or what kind of network is suitable using OSPF?" – Ron Maupin Dec 20 '20 at 23:46
  • @RonMaupin - I don't care about his identity. Yes privacy is a valid point. I just want to know his/her reason. We probably need a new feature where one can post a comment without revealing his ID (e.g. some tick box to hide ID). – caveman Dec 23 '20 at 13:45
  • I believe that has been proposed on Meta Stack Exchange a few times. I understand your frustration. I often get a vote down on an answer that has a dozen or more votes and has been accepted by the OP. It really makes you wonder what the voter was thinking, but, or course, everyone is entitled to an opinion and free to vote in any way they want. – Ron Maupin Dec 23 '20 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.