I have always wondered what causes an network engineer to pick EIGRP or OSPF over the other. Obviously OSPF is opened standard so can be used in a multi-vendor environment. If all equipment was cisco, what factors would you consider to implement OSPF and which factors would you consider for EIGRP?
To choose a routing protocol you should consider these topics:
Configuration: How easy is to configure and troubleshoot the protocol.
Efficiency: How fast it converges? How much bandwith it uses?
- Size and topology: Is the network a star? partial mesh? full mesh?
I recommend you to read this document about EIGRP vs OSPF.
I don't think that EIGRP is yet playing a role in today's routing networks. I have not seen any EIGRP Networks for about 5 years at my customer's networks. I think that Cisco will put EIGRP to a deprecated status within the next years.
The main reason i think is that there is an enormous amount of new players within the enterprise routing world that can compete with Cisco. Name them Juniper, Huawei and Nokia (formerly Alcatel Lucent). I have not talked of big players in Datacenter Layer3 Switching or NG-Firewall vendors that want to be implemented in today's routing structures too.
The document referenced in above Post is from 2007. Ten years ago there was Cisco and Cisco and nothing else, so i don't think that this article should be a reference for today's decision. On current platforms it is quite easy to implement an OSPF network that converges in milliseconds, even with single-Area implementations. Routers equiped with Hexacore Xeons do not start sweating on dijkstra's calculations like the poor MIPS boxes did 10 Years ago.
My scheme is: if you need a routing protocol to make routers reachable to each other (as a base for MPLS platforms or similar): use ISIS! If you want to make networks reachable by each other: use OSPF! If you want to interconnect networks: use BGP!
Follow the KISS principle when designing systems (including networks). The decision to introduce an interior dynamic routing protocol should be balanced against the overhead it introduces. Overhead meaning administrative effort, new admin handoff, maintenance (patches, sw updates, etc), interoperability, etc. I've even seen mid sized companies introduce BGP for all interior routing. They would spend weeks, months, fighting the most mundane of tasks. I won't name them but will say that particular client was one of the most respected names in I.T. ...Very perplexing.
Unfortunately, network engineers are often eager to enable features because of technical training. For example, great cooks know that the only seasoning a rib-eye should get is salt. Green cooks are gonna over do it with a concoction, or worse a marinade. Same idea with network engineers and most other arts.
I think protocol selection is just like selecting any product or service: your needs will lead you to the best solution. Each protocol has advantages and disadvantages when it comes to scalability, speed, reliability, security, etc.
If you are designing a network, you should consider the top factors for the owner of that network. Selecting the protocol is easy after that - there are many sources of information out there that detail and compare how each protocol addresses those factors.
Bottom line: The most important factor in selecting a protocol is making sure it best meets the customer's need.