IS-IS fundamentals: NET addresses, levels, DIS and pseudonode, metrics and configuration

Where IS-IS comes from

Intermediate System to Intermediate System (IS-IS) was designed for the OSI protocol suite (ISO/IEC 10589) and extended to carry IP routes (RFC 1195, "integrated IS-IS"). In OSI terms a router is an intermediate system and a host is an end system. Like OSPF it floods link-state information and runs SPF, but it differs in ways that suit large provider networks:

• It runs directly over Layer 2, not over IP, so it is harder to attack from a remote IP host and does not depend on IP being configured correctly.
• New information is carried in flexible TLV fields, which is why IPv6, traffic engineering and segment routing were easy to add.
• Its administrative distance on Cisco devices is 115.

NET addresses

Each IS-IS router is identified by a network entity title (NET). The part before the last seven bytes is the area; the next six bytes are the system ID, unique in the domain; the final byte (NSEL) is always 00 for a router. Addresses starting with 49 are private, like RFC 1918 addresses in IPv4. A router belongs to one area (unlike OSPF interfaces), so area borders fall on the links between routers.

Levels and packet types

Level 1 routing happens inside an area; level 2 routing connects areas and forms the backbone. A level 1/2 router does both, like an OSPF ABR. Level 1 routers do not learn other areas' routes; instead the level 1/2 router sets the attached (ATT) bit, and level 1 routers install a default route towards it. Cisco routers are level 1/2 by default; many providers simplify and run everything as level 2 only.

DIS and pseudonode

On a LAN, IS-IS elects a designated intermediate system (DIS), similar in purpose to the OSPF DR, with important differences. The highest interface priority wins (0 to 127, default 64), then the highest MAC address. The election is pre-emptive, so a better router takes over as soon as it appears. There is no backup DIS, and every router still forms an adjacency with every other router on the LAN.

The DIS creates a pseudonode: an imaginary router that represents the LAN in the database, so each real router only describes one link to the pseudonode instead of links to everyone. The DIS also sends CSNPs every 10 seconds so others can spot missing LSPs, and sends its hellos three times as often as other routers. On links with only two routers, set isis network point-to-point to skip the DIS entirely.

Metrics and configuration

Every interface has a default metric of 10, whatever its speed, so you must set metrics yourself to reflect bandwidth. Original narrow metrics allow only 63 per interface and 1023 per path. Wide metrics allow 24-bit interface and 32-bit path values and are needed for traffic engineering. Use metric-style wide on every router.

router isis CORE
 net 49.0001.0100.0000.0001.00
 is-type level-2-only
 metric-style wide
 log-adjacency-changes
 passive-interface Loopback0
!
interface GigabitEthernet0/0
 ip router isis CORE
 isis network point-to-point
 isis metric 100

Verify with show isis neighbors, show isis database, show clns interface and show ip route isis.

Real-world example: a regional fibre provider ran IS-IS level 2 across 60 routers. A new 100 Gbit/s link sat unused because, like every link, it had the default metric of 10, and the old path had fewer hops. The provider defined a metric standard (for example 100 for 10 Gbit/s and 10 for 100 Gbit/s), switched every router to wide metrics, and applied the standard. Traffic moved to the new link as intended.