Longest-Prefix Match, AD, and Metrics

When several routes could carry a packet, the router applies three tests in order and stops at the first one that decides. Get the order right and every route-selection question becomes mechanical.

1. Longest-prefix match — of all routes whose prefix contains the destination, the one with the most specific (longest) prefix wins. Decided first, and decided on the destination address alone.
2. Administrative distance — only if two or more routes match at the same longest prefix length but come from different sources, the source with the lower AD is installed.
3. Metric — only if the tie is still unresolved because two routes to the same prefix come from the same protocol, the lower metric wins.

There is a reason this order makes sense rather than being arbitrary. A more specific route reflects more precise knowledge about a destination: 10.1.1.0/24 says I know specifically about this /24, while 10.1.0.0/16 says I know about this larger block generally. Following the more specific route sends the packet toward the router with the finer-grained information, which is almost always the more correct path. The default route 0.0.0.0/0 is the logical endpoint of that idea — the least specific route, used only when nothing else matches.

R2# show ip route
Codes: C - connected, L - local, S - static, O - OSPF
       * - candidate default

Gateway of last resort is 203.0.113.1 to network 0.0.0.0

S*   0.0.0.0/0 [1/0] via 203.0.113.1
      10.0.0.0/8 is variably subnetted, 6 subnets, 4 masks
O        10.1.0.0/16 [110/20] via 10.9.9.1, 00:12:04, GigabitEthernet0/2
S        10.1.1.0/24 [1/0] via 10.9.9.5
C        10.1.1.0/26 is directly connected, GigabitEthernet0/0
L        10.1.1.1/32 is directly connected, GigabitEthernet0/0

Note that none of those three decisions consulted AD or metric. Prefix length settled all of them, which is what makes the first test so much more important than its position in a list suggests.

Two protocols can both learn a route to the exact same prefix — say OSPF and a static route both describe 10.1.5.0/24. Longest-prefix match cannot break that tie, since the prefix lengths are identical, so the router asks which source it trusts more. Lower AD wins.

So for the same prefix, a static route beats OSPF, which beats RIP. A connected route beats everything, which is why you never write a static route for a directly attached subnet. AD is a belief setting: it does not measure the path at all, it encodes how much the administrator trusts each source of information. That is exactly the mechanism a floating static route exploited in the previous lesson — raising a static route's AD above the primary's so it loses the comparison until the primary disappears.

If the surviving candidates are two routes to the same prefix from the same protocol — two OSPF paths, for instance — the router compares their metric and installs the lower. Each protocol computes its metric differently: OSPF uses a cost derived from bandwidth, EIGRP a composite of bandwidth and delay, RIP a hop count.

• Fixed order: longest-prefix match, then administrative distance, then metric — stop at the first test that decides.
• LPM first: the most specific matching prefix wins regardless of AD or metric.
• The default 0.0.0.0/0 is the least specific route and therefore the last resort.
• AD applies only to the same prefix from different sources: connected 0, static 1, EIGRP 90, OSPF 110, RIP 120, external EIGRP 170.
• Metric applies only within one protocol; comparing an OSPF cost to an EIGRP metric is meaningless.
• show ip route <destination> reveals the single entry that survived all three tests.