OSPF and IS-IS difference...

Dyasis

What is the difference? they both seem to do the same thing...

throw me a bone please?

Find more posts tagged with

Comments

NetworkVeteran

Dyasis wrote: »

What is the difference? they both seem to do the same thing...

throw me a bone please?

That's like Kindle vs. Nook, what's the difference? There's room in the router market for more than one approach.

IS-IS has more market share in the carrier space. OSPF has more market share in the enterprise space.

IS-IS was designed by the amazing Radia Perlman and went through extended review and standardization. IS-IS runs over L2 and was intended to support and be easily expanded to support multiple routed protocols--e.g., IPX, IPv4, and IPv6. It also is a bit more scalable, as an link/address change doesn't trigger a re-calculation. Its logic has even been incorporated into other protocols (TRILL)! The context that Cisco gave it more TLC early on, and there aren't any strong reasons to switch, also shouldn't be underestimated.

In comparison, OSPFv2 was developed and implemented quickly via the RFC process. The specification is easier to obtain and understand. OSPFv2 runs over IPv4 and routes for IPv4. Obviously, a simplified approach also has its flaws, especially when we consider the IPv6 migration, or incorporation into other protocols! I've seen enterprises run OSPFv2 for IPv4 and a separate protocol for IPv6. OSPFv3 was much more than a port to IPv6, however, and incorporates better scalability features plus IPSEC authentication.

Mrock4

NetworkVeteran wrote: »

Its logic has even been incorporated into other protocols (TRILL)!

And Cisco FabricPath- I know they're close, but I always see TRILL and think FabricPath. Old cisco habits! It's really trippy to think a "routing protocol" is used for Layer 2 loop prevention.

Dyasis

Ok, thanks for the replies

This stuff gets more and more interesting the deeper I go lol

powmia

Juniper also uses IS-IS in their Virtual Chassis implementation.

In addition to the scalability that IS-IS has by treating prefixes as leaf nodes:

IS-IS is also more extensible. IS-IS uses a TLV hierarchy to add new functionality to the protocol, as opposed to OSPFv2 using opaque LSAs. With the example of MPLS-TE, OSPFv2 uses opaque LSAs to carry CSPF info... which have been known to wreak havoc on things like mainframes that are listening to OSPF, but don't understand them. This causes people to create a multi-area topology to isolate parts of their network from what is running in the core.

IS-IS is more lightweight on the DR implementation for broadcast segments.

OSPF areas can be thought of as a collection of routers, with a router acting as the border between two areas.

IS-IS domains (levels) are a collection of links, with a L1-L2 link acting as the border between domains. Makes IS-IS much more flexible in the extension of backbone levels.

OSPF and IS-IS have a completely opposite approach to their default multi-area/multi-level architecture. In OSPF... with a two level hierarchy, using normal areas, all prefix information is sent to non-backbone areas. You have to modify the behavior (change area types) to limit the information sent to non-backbone areas. In IS-IS... with a two level hierarchy, by default, no information is sent to non-backbone levels (only a flag that tells nodes to create a default route). You have to modify the behavior to add to the information sent to non-backbone areas.

Now... OSPFv3 is a different story. They took some lessons learned from IS-IS. Prefix information is now separate from the router LSAs, by introducing the type 0x9 LSA (intra-area-prefix LSA)... now OSPFv3 behaves just like IS-IS in that regard. Also, OSPFv3 got rid of opaque LSAs and introduced a proper TLV hierarchy.

One last difference, is a comparison if IS-IS and OSPFv3. OSPFv3 can now support both IPv6 and IPv4, but it does so by creating a different topology for each. IS-IS, by default (and used to only support this) uses a single topology for both IPv4 and IPv6. This created the possibility of having a node somewhere in your topology that only had a single protocol (say, IPv4) running on a link that was calculated as the shortest path to the destination of an IPv6 prefix... causing traffic to blackhole. They then came out with an RFC for M-ISIS (Multi-Topology IS-IS), which you have to explicitly enable.