Sunday, March 4, 2012

IS-IS Implementations on Broadcast Networks and Point-to-Point Links

There are 2 general types of network topologies:
Broadcast networks LAN links (eg: Ethernet, Token Ring, Fiber Distributed Data Interface – FDDI) and multipoint WAN links.
Point-to-Point links Permanently established (eg: leased line, permanent virtual circuit – PVC) and dynamically established (eg: ISDN, switched virtual circuit – SVC) point-to-point links.

IS-IS only supports the following 2 modes for its link-state information.
There are no commands to change the network type as with OSPF.
Broadcast Default for LAN links and multipoint WAN links.
Broadcast mode is recommended for use only on LAN interfaces.
Point-to-Point Default for all other media types, including point-to-point subinterfaces and dialer interfaces.
Note: Avoid implementing IS-IS on dialup connections that incur usage-based costs, as IS-IS does not implement the Demand Circuit extension as in OSPF which may make such connections to stay up permanently and unwanted billing costs due to periodic IIHs.

IS-IS has no concept about NBMA. It is highly recommended to implement point-to-point links instead of multipoint links in NBMA environments (eg: X.25, ATM, or Frame Relay). IS-IS has no specific support for NBMA networks. When implementing a NBMA in broadcast mode using the broadcast keyword in the static DLCI mapping commands, Cisco IOS assumes that there is a NBMA network with full-mesh PVCs and able to advertise multicast updates. Static CLNS maps must also be created in addition to the static IP maps using the frame-relay map clns {dlci-number} broadcast interface subcommand.

2 IS-IS routers must support the same level of routing (L1 or L2) in order to form an adjacency. Separate IS-IS adjacencies are established for each level of routing. 2 neighboring routers on the same area perform both L1 and L2 routing would establish both L1 and L2 adjacencies. An IS-IS router maintains the L1 and L2 adjacencies in separate L1 and L2 adjacency tables. IS-IS routers on a LAN establish L1 and L2 adjacencies with all other routers on the LAN using specific L1 and L2 IIHs; while OSPF routers on a LAN establish FULL adjacencies only with the Designated Router (DR) and Backup Designated Router (BDR).

IIHs announce the area address, and the L1 and L2 neighbors of the originating routers. An adjacency is formed when the area address and the IS type as communicated via the IIHs are matched. L1 routers accept L1 IIH PDUs from their own area and establish adjacencies with other routers in their own area. L2 routers (and also the L2 process within an L1/L2 router) accept only L2 IIH PDUs and establish only L2 adjacencies. Unlike OSPF, the Hello intervals and holding time between 2 IS-IS neighboring routers do not need to be matched.

IIHs are padded to the full MTU size, which allows early error detection due to transmission problems with large frames or errors caused by mismatched MTUs on adjacent interfaces. Hello padding can be disabled in order to conserve network bandwidth in case the MTU of both interfaces is the same or translational bridging. When hello padding is disabled, Cisco routers still send the first 5 IIHs padded to the full MTU size in order to discover MTU mismatches. The no hello padding IS-IS router subcommand and the no isis hello padding interface subcommand disable hello padding for all interfaces and a particular interface respectively.

In IS-IS, a broadcast LAN is modeled as a directed graph or digraph with all the attached routers connected to a virtual router or pseudonode in a star topology manner. The pseudonode is actually the designated router or DIS (Designated Intermediate System) of the LAN. The virtual router or pseudonode makes the broadcast medium appeared as a virtual router and the routers appeared as its connected interfaces. It is responsible for generating LSPs on behalf of the LAN upon changes of its connections, eg: when a new neighbor comes online or offline. All routers maintain adjacencies to only the pseudonode instead of all routers on the LAN; thus reducing memory, CPU, and bandwidth resources. The adjacencies are managed by the DIS.

An IS-IS router on a LAN establish adjacencies with all other routers (including the DIS) through the pseudonode. Each router (including the DIS) establishes a single adjacency to the pseudonode rather than having each router establishes an adjacency with every router on the LAN. Otherwise, there are n x 2 adjacencies established on broadcast network with n connected routers, and each router would required to establish n adjacencies to every router; nevertheless, generating LSPs for every adjacency during LSDB synchronization creates considerable overhead!

A pseudonode is simply a virtual router; a real router must perform the tasks of the pseudonode. The DIS of the LAN takes on the responsibilities of the pseudonode which includes creating and maintaining adjacencies with all routers on the LAN, creating and updating the pseudonode LSP, and flooding LSPs over the LAN. The DIS sends out separate L1 and L2 LSPs for the pseudonode.

The criteria for selecting the DIS are the router highest priority followed by the highest SNPA (the SNPA on LANs is the MAC address). Cisco router interfaces have a default L1 and L2 priority of 64. The priority value from 0 to 127 can be configured for L1 and L2 independently using the isis priority {priority-value} [level-1 | level-2] interface subcommand. The L1 DIS and L2 DIS on a LAN may or may not be the same router, as an interface can have different L1 and L2 priorities. Setting the priority to 0 only lowers the chance of a router to become the DIS, but does not prevent it. When a router with a higher priority is being introduced to the LAN, it will preempt and take over the DIS role from the current DIS (different than OSPF). Since the IS-IS LSDB is synchronized frequently on a LAN (every 10 seconds), handing over the DIS role to another router is not a significant issue.

When the current DIS fails, another router would take over and become the new DIS instantly with little or no impact upon the network. There is no backup designated router or DIS in IS-IS. Contrast this behavior with OSPF, where the DR and BDR are selected and the other routers on the LAN establish FULL adjacencies only with the DR and BDR. In case of DR failure, the BDR is promoted to become DR, and a new BDR is elected.

IS-IS Adjacencies over a Broadcast Network

Figure above shows the IS-IS adjacencies over a broadcast network and how the DIS generates the Pseudonode LSPs. A pseudonode LSP details only the adjacent router connected to the LAN. The Pseudonode is logically connected to all routers; all routers still establish adjacencies among themselves. The Pseudonode LSP is used to build the network map and eventually the SPT. The Pseudonode LSP is the equivalent of a Type-2 Network-LSA in OSPF.
Note: The DIS doesn’t actually establish adjacencies and synchronize LSDB with all routers; it is the Pseudonode, a virtual router that created by the DIS.

IS-IS uses a 2-level area hierarchical. The link-state information for the 2 levels is distributed separately using L1 and L2 LSPs. Each IS originates its own LSPs (one for L1 and one for L2). L1 and L2 IS-IS LAN PDUs are sent periodically as multicasts using multicast MAC addresses. L1 and L2 IIHs, LSPs, and SNPs on a LAN are sent to the AllL1IS multicast MAC address 0180.C200.0014 and the AllL2IS multicast MAC address 0180.C200.0015 respectively.
IS-IS PDUs are sent out as multicasts on broadcast networks; and as unicasts on point-to-point links.

IIHs are used to establish and maintain adjacencies between routers. If a router does not receive an IIH from a neighboring router within the holding time, the neighboring router is declared dead and all routing entries associated with the router are removed from the routing table. Note that the database entries associated with the router still remain in the link-state database. The holding time is calculated as the product of the Hello multiplier and Hello interval. The default Hello interval is 10 seconds and the default Hello multiplier is 3; therefore the default holding time is 30 seconds. The Hello interval can be adjusted using the isis hello-interval {sec} [level-1 | level-2] interface subcommand. Unlike OSPF, the Hello intervals and holding time between 2 IS-IS neighboring routers do not need to be matched.

The IS-IS adjacencies on a LAN is maintained by the DIS. The DIS sends out Hellos 3 times faster than the Hello interval of other routers – 3.3 seconds, in order to detect DIS failure quickly.

When a network consists of only 2 IS-IS routers over a broadcast network, the connection can be treated as a point-to-point link instead if a broadcast network.

Unlike LAN interfaces which generate separate L1 and L2 IIHs, point-to-point links have a common point-to-point IIH format that specifies whether the Hello is for L1 or L2 or both. Point-to-point IIHs are sent to the unicast address of the neighboring router at the other end.

Below summarizes the differences between IS-IS broadcast and point-to-point modes:

Broadcast Mode Point-to-Point Mode
Usage LANs and full-mesh WANs. PPP, HDLC, and partial-mesh WANs.
Hello interval 3.3 seconds for DIS;
10 seconds for others.
10 seconds
Adjacencies n x 2 n – 1
Uses DIS? Yes No
IIH Type L1 IIH and L2 IIH Point-to-Point IIH

3 comments:

  1. Wow very informative post. I have read full post and it is very effective. Thanks for your post.
    wan leased line

    ReplyDelete
  2. Very informative post.A lot of thanks for your post.
    point to point

    ReplyDelete
  3. Thanks for the post. Very good job!
    Nevertheless, I have a question Yap Chin Hoong, I'm not able to understand if the adjacencies are stablished between routers on a LAN or only between routers attached to LAN and the DIS. The picture depicts this fact in the "note", but you say "Each router (including the DIS) establishes a single adjacency to the pseudonode rather than having each router establishes an adjacency with every router on the LAN".

    ReplyDelete