Home » Bury the hatchet » IS-IS and fast convergence ongoing tricks

IS-IS and fast convergence ongoing tricks

Been a while since my last post, I was extremely busy doing a lot of things, anyway I am glade to be back.

This post I am going to cover a nice tool for enhancing IS-IS convergence, I am really amazed by the ideas that the guys out there pop up. Inventing such wonderful tools requires intellectual open minds (it’s not relatively a new feature), enough talking and lets get to the point.

Practically speaking in most Service Providers’ backbones the IGP is required to make the loopback IP addresses of the providers’ routers reachable to each other in order to be able to have all the MPLS stuff operational, no other IP addresses information is required to be exchanged by the IGP in the global routing table for things to work out (if the Service Provider is providing internet through the global routing table, this kind of routes are carried via BGP and not the IGP).

IS-IS has a characteristic that has always made it more flexible and scalable (at least in the single area designs) than OSPF, and that is it separates the neighbors and the IP addresses information. It uses IS Reachability TLV (or Extended IS Reachability TLV with wide metrics) and IP Internal/External Reachability Information TLV (or Extended IP Reachability TLV with wide metrics).

This well engineered data-structure of IS-IS has leveraged it huge benefits when it came to Partial SPF calculation (AKA Partial Route Computation – PRC) , since it was easy for a router to understand whether there is a topology change that needs full SPF calculation, or is it just IP addresses reachability that would just require a partial SPF calculation.

This well engineered data-structure keeps providing IS-IS with flexibility that features like the one that I am going to talk about today leverages as well. Cisco names the feature “IS-IS Mechanism to Exclude Connected IP Prefix from LSP Advertisements”. It is all about minimizing the database that is required to be flooded, synchronized, stored and processed on all the network routers and thus enhancing the convergence process.

Back to the second paragraph, since in most service providers’ backbones only the looback IP addresses reachability is required for proper operation (I am of course talking about a MPLS backbone), thus why not only maintain such information and discard any other useless information.

Cisco has introduced two mechanisms for this feature (check Cisco’s documentation):

  • Explicit (Small-Scale method as named by Cisco): In this method you simply explicitly configure an IS-IS interface not to advertise its IP information to neighbors using the no isis advertise-prefix interface command – Which is of course not a scalable method (that’s how it got its name).
  • Passive Only (Large-Scale method as named by Cisco): In this method you simply notify the IS-IS router not to advertise any connected IP information to its neighbors using the advertise passive-only command under the IS-IS process, except those configured as passive-interfaces, which is the most common way of advertising the loopback interfaces.

The logic behind this feature is extremely simple; “don’t advertise useless IP addresses reachability information” by not advertising their IP reachability TLVs in the LSPs.

This feature affects only connected interfaces running IS-IS, it has nothing to do with redistribution, and accordingly any redistributed IP information is never affected by this information.

Find below the lab illustration of this feature – this is my fast lab with six inline routers: CE1(R5) <–> PE1(R1) <–> P1(R2) <–> P2(R3) <–> PE2(R4) <–> CE2(R6).

Note that in the first output both the connected IP information (12.12.12.0/24) and the redistributed IP information (111.111.111.111/32) are received by P2, while in the second output, after configuring this feature on PE1 only the connected IP information disappears, while the redistributed IP information persists, since this command excludes the connected IP prefixes only.

Before configuring the feature on PE1

P2(R3)#sh isis database verbose 

IS-IS Level-2 Link State Database:
LSPID                 LSP Seq Num  LSP Checksum  LSP Holdtime      ATT/P/OL
PE1(R1).00-00         0x000002FD   0x6CF8        1191              0/0/0
  Area Address: 49.0001
  NLPID:        0xCC
  Hostname: PE1(R1)
  IP Address:   1.1.1.1
  Metric: 10         IS-Extended P1(R2).00
  Metric: 0          IP 1.1.1.1/32
  Metric: 10         IP 12.12.12.0/24
  Metric: 0          IP 111.111.111.111/32
P1(R2).00-00          0x000002FE   0x5509        515               0/0/0
  Area Address: 49.0001
  NLPID:        0xCC
  Hostname: P1(R2)
  IP Address:   2.2.2.2
  Metric: 10         IS-Extended PE1(R1).00
  Metric: 10         IS-Extended P2(R3).01
  Metric: 0          IP 2.2.2.2/32
  Metric: 10         IP 12.12.12.0/24
  Metric: 10         IP 23.23.23.0/24

After configuring the feature on PE1

P2(R3)#sh isis database verbose 

IS-IS Level-2 Link State Database:
LSPID                 LSP Seq Num  LSP Checksum  LSP Holdtime      ATT/P/OL
PE1(R1).00-00         0x000002FE   0xB6FB        1183              0/0/0
  Area Address: 49.0001
  NLPID:        0xCC
  Hostname: PE1(R1)
  IP Address:   1.1.1.1
  Metric: 10         IS-Extended P1(R2).00
  Metric: 0          IP 1.1.1.1/32
  Metric: 0          IP 111.111.111.111/32
P1(R2).00-00          0x000002FE   0x5509        474               0/0/0
  Area Address: 49.0001
  NLPID:        0xCC
  Hostname: P1(R2)
  IP Address:   2.2.2.2
  Metric: 10         IS-Extended PE1(R1).00
  Metric: 10         IS-Extended P2(R3).01
  Metric: 0          IP 2.2.2.2/32
  Metric: 10         IP 12.12.12.0/24
  Metric: 10         IP 23.23.23.0/24

I hope that I’ve been informative.

BR,
Mohammed Mahmoud.

6 comments

  1. Could you discuss the convergence alternatives
    in this case i.e when connected information is not
    advertised. Is it going to be purely MPLS re-route
    schemes. It would also impose restrictions on
    usagee of MPLS forwarding adjacencies.
    THanks

  2. Hi,

    I am not sure that I’ve understood you thoroughly. Actually when IS-IS stops sending the connected IP prefixes in the LSPs, the network routers won’t be having these IP addresses in their routing table, and thus their label binding won’t be populated into the LFIB, thus we are now left only with MPLS traffic between the loopback IP addresses, which is enough for a MPLS based Service Provider (if any other traffic flow requires to be label switched for any reason, you can still manipulate things and make it work).

    Regarding MPLS re-route, I believe that using MPLS TE and FRR is a crucial tool for a MPLS based Service Provider, it provides a flexible, deterministic and well controlled network – If your point was the interaction of the described feature with MPLS TE, I haven’t tried this out (needs to be labed), but from a higher design level, generally when using MPLS TE don’t bother tuning the IGP.

    For an overall good design you’ll have to evaluate all the tools you have, which of them is supported on the platforms and IOS codes you are running, the pros and cons for each of them in your particular network topology and finally lab things until you completely feel the behavior.

    I hope that I’ve been informative.

    BR,
    Mohammed Mahmoud.

  3. Hi,
    I was just
    pointing out the fact that IGP convergence mechanism
    is no longer relevant and networks will have to rely
    on pure MPLS-TE schemes. Having said that, how would it impact CSPF and the TED database if ISIS does not provide
    link information.

    regards

  4. Hi,

    I must admit that you are raising very interesting points, I really enjoy this discussion.

    Like I’ve said in my first comment you’ll have to evaluate your needs, tools and resources to reach what best fits your design, there is no one answer to the question of which is the best, tuning IGP convergence might be the best approach in some circumstances/networks while in others it’s not. Generally yes relying on MPLS-TE should be the best approach for large scale networks and in such case tuning the IGP would be of minor need, but again I don’t admire generalizing.

    For the second point, after doing some testing in the lab, what really happens is that this command stops IS-IS from exchanging the IP reachability information, but this doesn’t intervene by any mean with the exchange of the traffic engineering extensions, and thus the shortest unconstrained path calculation is not affected – you can do a simple lab to test this using three routers in-line and use “debug isis mpls traffic-eng advertisements” and “debug isis mpls traffic-eng events”.

    I hope that I’ve been informative.

    BR,
    Mohammed Mahmoud.

  5. i want really . why is is-is more flexible and stable than ospf ? give me examples?

  6. For my last project I had to study this comparison
    ISIS vs OSPF. I had the chance to compare two different famous network equipments vendor. Finally I have understood that it’s really dependant of what you are building and for who (SP, Corpo etc …). In addition to that, note that it’s also interesting to focus on the services you have to setup L3vpn, L2vpn, TE, etc… . In our case ISIS was more adapted since we have to setup L2vpn and that TRILL (a new protocol that will be very usefull) is developped for ISIS.

    Kamel

Leave a Reply

Your email address will not be published. Required fields are marked *