Advanced Routing Services (OSPF and RIP)

In addition to Policy Based Routing and RIP advertising, SonicOS Enhanced offers the option of enabling Advanced Routing Services (ARS). Advanced Routing Services provides full advertising and listening support for the Routing Information Protocol (RIPv1 - RFC1058) and (RIPv2 - RFC2453), and Open Shortest Path First (OSPFv2 – RFC2328). Advanced Routing Service should only be enabled by those environments requiring support for either or both of these dynamic routing protocols.

RIP and OSPF are Interior Gateway Protocols (IGP) that are both widely used by networks of various sizes to automate the process of route distribution. RIP is commonly used within smaller networks, while OSPF is used by larger networks, although network size should not be the only factor used to determine the appropriateness of one protocol over the other – network speed, interoperability requirements, and relative overall complexity, for example, should also be considered. RIPv1 and RIPv2 are both supported by ARS, the largest differences between the two being that RIPv2 supports VLSM (Variable Length Subnet Masks), authentication, and routing updates. The following table illustrates the major differences between RIPv1, RIPv2, and OSPFv2:

Routing Information Protocol Differences
	RIPv1	RIPv2	OSPFv2
Protocol Metrics	Distance Vector	Distance Vector	Link State
Maximum Hops	15	15	Unlimited
Routing Table Updates	Full table broadcast periodically, slower convergence	Full table broadcast or multicast periodically, slower convergence	Link state advertisement multicasts, triggered by changes, fast convergence
Subnet Sizes Supported	Only class-based (a/b/c) subnets support	Class-based only	VLSM
Autonomous System Topology	Indivisible and flat	Indivisible and flat	Area based, allowing for segmentation and aggregation

•

Protocol Metrics – Distance Vector protocols such as RIP base routing metrics exclusively on hop counts, while Link state protocols such as OSPF consider the state of the link when determining metrics. For example, OSPF determines interface metrics by dividing its reference bandwidth (100mbits by default) by the interface speed – the faster the link, the lower the cost and the more preferable the path. Consider the following example network:

Sample network

In the above sample network, if Host A wanted to reach Host B, with RIP, the lowest cost route would be from Router A to Router B, across the relatively slow 64kbps link. With OSPF, the cost from Router A to Router B would be 1562, while the cost from Router A to Router C to Router D to Router B would be 364 (see the Cost section in OSPF concepts later), making it the preferred route.

•

Maximum Hops – RIP imposes a hop count of 15 to help prevent routing loops which can occur when bad (for example, stale) routing information is broadcast and propagated through a network either due to misconfiguration, or slow convergence. Consider if the link between Router D and Router E failed in the diagram above, and there were no safeguards in place:

•

Router A’s routing information states that it can reach Network E through Router B or Router C with a metric of 3.

•

When the link between Router D and Router E fail, and Router A broadcasts its routing information, Router B and Router C determine that they can reach Network E through Router A with a metric of 4.

•

Router B and Router C broadcast this information, and it is received by Router D which then determines it can reach Network E through Router B or Router C with a metric of 5.

•

This loop continues until the hop count of 16 (infinity) is reached.

Other measures against this sort of situation are also commonly employed by RIP, including:

•

Split-Horizon – A preventative mechanism where routing information learned through an interface is not sent back out the same interface. This generally works well on broadcast links, but not on non-broadcast links such as Frame Relay, where a single link can commonly be used to reach two separate autonomous systems.

•

Poison reverse – Also known as route poisoning, an extension of split-horizon where a network is advertised with a metric of 16 (unreachable), helping to ensure that incorrect alternative routes aren’t propagated.

OSPF does not have to impose a hop count limit because it does not advertise entire routing tables, rather it generally only sends link state updates when changes occur. This is a significant advantage in larger networks in that it converges more quickly, produces less update traffic, and supports an unlimited number of hops.

•

Routing table updates – As mentioned above, the practice of sending an entire routing table introduces the problems of slower convergences, higher bandwidth utilization, and increased potential for stale routing information. RIPv1 broadcasts its entire routing table at a prescribed interval (usually every 30 seconds), RIPv2 can either broadcast or multicast, and OSPF multicasts only link state updates whenever a change to the network fabric occurs. OSPF has a further advantage of using designated routers (DR) in forming adjacencies in multiple-access networks (more on these concepts later) so that updates do not have to be sent to the entire network.

•

Subnet sizes supported – RIPv1 was first implemented when networks were strictly class A, class B, and class C (and later D and E):

•

Class A – 1.0.0.0 to 126.0.0.0 (0.0.0.0 and 127.0.0.0 are reserved)

•

Leftmost bit 0; 7 network bits; 24 host bits

•

0nnnnnnn hhhhhhhh hhhhhhhh hhhhhhhh (8-bit classful netmask)

•

126 Class A networks, 16,777,214 hosts each

•

Class B - 128.0.0.0 to 191.255.0.0

•

Leftmost bits 10; 14 network bits; 16 host bits

•

10nnnnnn nnnnnnnn hhhhhhhh hhhhhhhh (16-bit classful netmask)

•

16,384 Class B networks, 65,532 hosts each

•

Class C – 192.0.0.0 to 223.255.255.0

•

Leftmost bits 110; 21 network bits; 8 host bits

•

110nnnnn nnnnnnnn nnnnnnnn hhhhhhhh (24-bit classful netmask)

•

2,097,152 Class Cs networks, 254 hosts each

•

Class D - 225.0.0.0 to 239.255.255.255 (multicast)

•

Leftmost bits 1110; 28 multicast address bits

•

1110mmmm mmmmmmmm mmmmmmmm mmmmmmmm

•

Class E - 240.0.0.0 to 255.255.255.255 (reserved)

•

Leftmost bits 1111; 28 reserved address bits

•

1111rrrr rrrrrrrr rrrrrrrr rrrrrrrr

This method of address allocation proved to be very inefficient because it provided no flexibility, neither in the way of segmentation (subnetting) or aggregation (supernetting, or CIDR – classless inter-domain routing) by means of VLSM – variable length subnet masks.

VLSM, supported by RIPv2 and OSPF, allows for classless representation of networks to break larger networks into smaller networks:

For example, take the classful 10.0.0.0/8 network, and assign it a /24 netmask. This subnetting allocates an additional 16-bits from the host range to the network range (24-8=16). To calculate the number of additional networks this subnetting provides, raise 2 to the number of additional bits: 2^16=65,536. Thus, rather than having a single network with 16.7 million hosts (usually more than most LAN’s require) it is possible to have 65,536 networks, each with 254 usable hosts.

VLSM also allows for route aggregation (CIDR):

For example, if you had 8 class C networks: 192.168.0.0/24 through 192.168.7.0/24, rather than having to have a separate route statement to each of them, it would be possible to provide a single route to 192.168.0.0/21 which would encompass them all.

This ability, in addition to providing more efficient and flexible allocation of IP address space, also allows routing tables and routing updates to be kept smaller.

•

Autonomous system topologies – An autonomous system (AS) is a collection of routers that are under common administrative control, and that share the same routing characteristics. When a group of autonomous systems share routing information, they are commonly referred to as a confederation of autonomous systems. (RFC1930 and RFC975 address these concepts in much greater detail). In simple terms, an AS is a logical distinction that encompasses physical network elements based on the commonness of their configurations.

With regard to RIP and OSPF, RIP autonomous systems cannot be segmented, and all routing information must be advertised (broadcast) through the entire AS. This can become difficult to manage and can result in excessive routing information traffic. OSPF, on the other hand, employs the concept of Areas, and allows for logically, manageable segmentation to control the sharing of information within an AS. OSPF areas begin with the backbone area (area 0 or 0.0.0.0), and all other areas must connect to this backbone area (although there are exceptions). This ability to segment the routing AS helps to ensure that it never becomes too large to manage, or too computationally intensive for the routers to handle.