Saturday, December 24, 2011

Interconnecting Multiple OSPF Areas

In large networks which the web of router links is often very complex due to the large number of potential paths to all subnets, the SPF calculations that considering all the possible paths would be very complex and consume significant time and router resources.

OSPF is able to reduce the effort of SPF calculations by partitioning a network into areas. With smaller number of routers and LSAs that flood within an area, the link-state database would become smaller, in which the SPF calculations are relatively easier and faster. Additionally, the flooding of information upon a router or link failure only affects the routers within an area; routers outside the area will not receive the info hence do not need to perform SPF calculation. Hierarchical routing is a often being used when building large networks.

OSPF implements a 2-level hierarchy routing scheme which comprises of the following area types:

Backbone or Transit area Provides fast and efficient forwarding of IP packets between standard areas. Backbone area distributes routing information between standards areas. End systems are not found within the backbone or transit area.
Standard or Regular area Connects end systems and resources. Standard areas (non-backbone areas) do not allow traffic from other areas to pass through it. All traffic from other areas must transmit across a transit area. It includes a number of subtypes, namely stub area, totally stubby area, and not-so-stubby (NSSA) area.
Furthermore, OSPF allows an autonomous system to be split into additional layers or levels by incorporating 2 levels of external routing (E1 and E2 external routes) into the OSPF routing domain.

LS routing protocols require all routers to maintain a copy of the LSDB; the more OSPF routers, the larger the LSDB. It would be nice to have all information of all routers, but this approach has scalability issues for larger networks. With the concept of area, routers inside an area only required to maintain detailed information about the routers and links within the area and summary information about routers and links in other areas. By implementing a hierarchical structure using areas, an OSPF autonomous system or routing domain can scale very large.
Note: Cisco recommends that generally there should be no more than 50 routers per area.

The Internet is split into autonomous systems with the boundaries of network segments. For OSPF, the routers are the area boundaries. This means that a network segment would belong to only one area, which eases the aggregation of network addresses across area boundaries.

The abstraction process summarizes a list of network addresses within an area when advertising the information of the area to other areas. The information is being passed from area to area, as like routes that are being advertised from router to router with the Distance Vector algorithm. Although the DV algorithm is used between areas, OSPF was designed to avoid the common pitfalls of DV algorithm by requiring that all areas must be directly connected to a special area known as the backbone areas, which results in a hub-and-spoke organization that DV algorithm is able to handle without any problem. The DV algorithm between areas will be discussed later.

The restriction that requires all areas to be directly connected to the backbone area could be compromised through the use of virtual links, which are logical extensions of the backbone area. Virtual links tunnel routing information through areas which create a logical hub-and-spoke area topology regardless of the physical area topology.

OSPF Router Types

Below describes the various types of OSPF routers:

Internal Router An OSPF router in which all its interfaces reside within a single area. All routers within the same area maintain the same LSDB for the area.
Backbone Router An OSPF router in which all its interfaces reside within the backbone area or has at least one interface resides in the backbone area.
Area Border Router (ABR) An OSPF router that has interfaces reside in the backbone area and non-backbone areas. An ABR maintains separate (complete) LSDBs for all areas that it resides, route packets between the areas, and separates the LSA flooding zones – LSA flooding would stop at the area boundary. An ABR is the primary point of routing information exchange for an area, distributes the routing information of a standard area into the backbone; the backbone routers propagate the info to other ABRs and eventually internal routers of other areas. An ABR can be configured to summarize / filter the routing information in its LSDB before sending it to other areas. In a multiarea OSPF network, an area can have multiple ABRs.
Autonomous System Boundary Router (ASBR) An OSPF router that has at least one interface connects to an external internetwork – another autonomous system or non-OSPF network. ASBRs can perform route redistribution to import / export non-OSPF network information between the OSPF routing domain and external internetwork.

Note: A router can be more than one router type. Ex: A router which connects to area 0, area 1, and a non-OSPF network is a backbone router, an ABR, and an ASBR.

Below are some OSPF design guidelines based on real-world studies and implementations:

Min Mean Max
Routers in a domain 20 510 1000
Routers in an area 20 160 350
Areas in a domain 1 23 60

i) An area should have no more than 50 routers.
ii) A router should have no more than 60 OSPF neighbors.
iii) A router should not be in more than 3 areas. The ideal design is to have an ABR connects to 2 areas only – the backbone and a standard area.
iv) A router can be a DR or BDR for more than 1 network segment.
v) Do not run more than 1 OSPF process on an ABR.

No comments:

Post a Comment