Question about IS-IS vs. OSPF

[mohd_arab]

i read in a cisco white paper a comparision between IS-IS and OSPF. the IS-IS was prefered because the core area in the IS-IS is extensible, the reason why is that?

because in IS-IS the area border lie "on links" not "inside router" as for OSPF. so in IS-IS you can add as many L2 router as you want to expand the core area.

wonderful!!

if the term "expansion of the core area means the number of routers inside the area so as fas as i know this could be accomplished by OSPF if we added more internal routers inside the core layer!!

"Am i right"??

If not, may someone explain it to me and what is exactly ment by "extension of core layer"??

[mohd_arab]

Sorry!!

*If not, may someone explain it to me and what is exactly ment by "extension of core area"??

[zander13]

I will try and provide my own interpretation here.

OSPF & IS-IS are very similair but IS-IS has the ability to scale to much larger size networks than OSPF.

OSPF only scales to 50-100 router per area. IS-IS on the other hand will scale to 1000 routers per area ! It is also harder to extend OSPF area 0 than a IS-IS core area.

So in principle you could have 1000 IS-IS routers in your core area where OSPF will only allow 50-100.

Why does IS-IS scale better than OSPF ? Because fewer LSPs are used etc which means less CPU/Memory overhead.

IS-IS TLVs also allow for more future extensions than OSPF LSAs.

IS-IS is a real beauty and very simple to implement. Because of it's simplicity it also has several shortcomings where OSPF shines. ie.

Hope this helps.

Cheers

mips

[tbaranski]

Do you have a link to the white paper?

[mohd_arab]

the link is:

http://www.ciscopress.com/articles/article.asp?p=31319&redir=1

its a cisco press article actually.

scroll down to the "Integrated IS-IS Versus OSPF" section. you will find it there

[zander13]

OSPF & IS-IS are very similair but IS-IS has the ability to scale to much larger size networks than OSPF.

OSPF only scales to 50-100 router per area. IS-IS on the other hand will scale to 1000 routers per area ! It is also harder to extend OSPF area 0 than a IS-IS core area.

So in principle you could have 1000 IS-IS routers in your core area where OSPF will only allow 50-100.

Why does IS-IS scale better than OSPF ? Because fewer LSPs are used etc which means less CPU/Memory overhead.

[mheusinger]

Hi,

I do not agree completely. Today also OSPF scales to similar numbers like IS-IS. Actually when talking to Cisco developers I got the statement, that both are very similar with respect to theoretical limits.

The OSPF limit of 50-100 routers is an old one. Really the limiting factors are memory and CPU on your routers. Tuned properly (area design, timers, summarization, ispf, etc.) both protocols perform up to several hundred routers per area. Personally I have seen OSPF networks with about 300 routers per area performing well.

regards

Martin

[mheusinger]

Second thoughts:

Is the main factor really the number of routers per area?

Choosing an IGP usually is influenced by a huge number of features: WAN/LAN performance, backup scenarios, ..., and not to forget existing hardware/software and protocol knowledge. (Controllers may add cost effectiveness ;-)

What is the "best" IGP is result of your personal preferences and needs.

And to answer your original question: Yes you are right, that is what is meant by "expansion of core area".

[ruwhite]

First, I'll agree--the IGP you choose is primarily a matter of which capabilities you want, versus what you're comfortable with, etc. In this case, a lot of confusion reigns between IS-IS and OSPF. There are differences, but they aren't what a pervious writer wrote--that IS-IS is simpler, therefore more limited.

In fact, the primary differences are, from a network design standpoint:

-- IS-IS and OSPF use different router ID's. Okay, big deal.... :-) Many people are intimidated by the net number, but it's really no different than configuring a router ID on each OSPF router in the network (which is what I would do, if I were deploying OSPF, anyway).

-- OSPF has a bit more in the way of troubleshooting capabilities, in terms fo debugs and such.

-- On the other hand, I find IS-IS' database easier to understand, so I find it easier to find problems in an IS-IS network. Ironic, I know, but there it is.

-- OSPF has a lot of flexibility in the way it handles point-to-multipoint links, next hops to externals, and some other corner cases that IS-IS doesn't have.

-- IS-IS has a _much_ more flexible hierarchy scheme, and not just because the L1/L2 border is on the link, rather than in the router. The L2 domain can actually overlap various L1 domains, making a lot of network designs possible in IS-IS that just aren't possible in OSPF.

-- OSPF, because of its complex LSA type stuff, tends to be a bit more efficient on the wire (or at least it should be in theory).

-- IS-IS tends to be easier to extend. For instance, IS-IS for IPv6 just defines a new set of TLVs. OSPF for IPv6 defines an entirely new protocol. IS-IS has one LSP type, with different TLV's, and some divisions in its metric space. OSPF has 12 LSA types, so far, if I've not missed a few, and counting.

All in all, it really depends on what you're more comfortable with, and if there are specific features on the list that you really need in either direction.

:-)

Russ.W

[mark-obrien]

I read that the IS-IS process must shut down for 20 minutes if the sequence number of an LSP counts down to zero. Is this true? I'd hate to have to explain that to the boss.

[tbaranski]

Russ,

"IS-IS has a _much_ more flexible hierarchy scheme, and not just because the L1/L2 border is on the link, rather than in the router."

I'm assuming you meant the area border. But regardless, I'm unclear on how the location of area or level borders affects protocol scalability. The document linked above (which seems to have a pro-ISIS bias to it, unless I'm imagining things) mentions lower LSP use, but I'm not making the connection. If you have any insights you'd like to share, I'd be appreciative.

[zander13]

If you have the course notes refer to BSCI v2.0 Chap 5, Pg 19. It clearly states the advantages of IS-IS over OSPF according to *Cisco*.

Would I be violating Cisco Copyright if I quoted the page ad verbatim here ?

I will summarise in the meantime.

Between Access & ABP (L1 & L1/L2):

OSPF produces many small advertisements, Router LSA, Network LSA, Summary LSA, External LSA etc. Each packet is routed (which uses resources) thus more packets have a larger impact on the network.

IS-IS updates are grouped and sent by the router as one LSP, packet size will increase with network complexity.

OSPF runs on top of IP, IS-IS uses ClNS which is encapsulates into the Data-link layer. Again less overhead.

IS-IS is less demanding on CPU resources based on the way it handles routing updates. There are fewer LSPs to process & the way networks are added & removed are less intensive because NET addresses are already heavily summarised.

When using DEFAULT timers IS-IS detects failures faster than OSPF, thus converges faster than OSPF. IS-IS permits more tuning of timers (there are more timers to adjust ?) than OSPF, increasing convergence even more. Keep in mind that tuning timers on both can affect stability.

New ideas require the creation of new LSAs in OSPF, with IS-IS TLVs are flexible and can incorporate new ideas easily as mentioned above.

Cheers

mips