5.1 Enterprise Routing Principles: Distance Vector vs. Link-State, EIGRP, and OSPF

Enterprise Routing Principles: Distance Vector vs. Link-State, EIGRP, and OSPF

Enterprise IP routing forms the bedrock of campus, data center, and wide-area network communications. Routers build forwarding paths by exchanging reachability information through Interior Gateway Protocols (IGPs). To optimize network convergence, avoid forwarding loops, and make deterministic traffic engineering decisions, network engineers must master the foundational algorithms, protocol structures, and route selection criteria that govern modern enterprise routing engines.

---

The Route Selection Hierarchy

When a router receives a packet, its forwarding plane inspects the destination IP address and queries the routing table (Routing Information Base, or RIB) to locate the best exit interface and next hop. If multiple routing protocols or static sources provide reachability data, the routing engine evaluates paths using a strict three-tiered decision hierarchy:

+-------------------------------------------------------------+
| 1. LONGEST PREFIX MATCH (LPM)                               |
|    Evaluates subnet mask length (/32 down to /0).           |
|    Always takes absolute precedence over all other criteria.|
+-------------------------------------------------------------+
                              |
                              v
+-------------------------------------------------------------+
| 2. ADMINISTRATIVE DISTANCE (AD)                             |
|    Evaluated ONLY when identical prefix/mask entries exist. |
|    Prefers the routing source with the lowest AD value.     |
+-------------------------------------------------------------+
                              |
                              v
+-------------------------------------------------------------+
| 3. PROTOCOL METRIC                                          |
|    Evaluated within the SAME protocol and prefix.           |
|    Lowest metric (cost, composite, hop count) is installed. |
+-------------------------------------------------------------+

1. Longest Prefix Match (LPM)

The router compares the destination IP address against all routing entries bit-by-bit from left to right. The entry with the most matching leading bits (the longest prefix length or most specific subnet mask) wins unconditionally. For instance, a /28 route is chosen over a /24 or /16 route regardless of which protocol installed them or what their administrative distance values are.

2. Administrative Distance (AD)

Administrative Distance is a local trustworthiness rating assigned to each routing information source. If two different routing protocols (or a static route and an IGP) advertise the exact same IP prefix and prefix length (e.g., both advertise 10.10.10.0/24), the router installs the prefix from the source with the lower AD value into the forwarding table.

3. Metric Comparison

If a single routing protocol learns multiple paths to the exact same prefix with identical subnet masks, it compares the protocol-specific metrics (such as OSPF cost or EIGRP composite metric). The path with the lowest metric is installed. If metrics are identical, Equal-Cost Multi-Path (ECMP) load balancing occurs across the matching paths.

---

Administrative Distance Hierarchy

Cisco IOS assigns default Administrative Distance values across routing mechanisms, ranked from most trusted (lowest number) to untrusted (highest number):

Floating Static Routes: By manually configuring a static route with an administrative distance higher than the active IGP (e.g., ip route 10.0.0.0 255.0.0.0 192.0.2.1 115 to back up OSPF at AD 110), the static route remains dormant in configuration until the primary IGP route fails, offering deterministic backup connectivity.

---

Interior Gateway Protocol (IGP) Architectural Comparison

Interior Gateway Protocols operate within a single autonomous domain and fall into three primary architectural classes: traditional Distance Vector, Link-State, and Advanced (Hybrid) Distance Vector.

---

Distance Vector vs. Link-State Mechanics

Distance Vector and "Routing by Rumor"

Traditional distance vector protocols determine the distance (metric) and vector (next hop direction) to target networks based solely on the second-hand advertisements of their direct neighbors. Routers possess no end-to-end topological awareness. Because routers depend on periodic broadcasts, distance vector networks are vulnerable to routing loops and the "count-to-infinity" problem when links flap.

Mitigation techniques in distance vector include:

• Split Horizon: A router never advertises a route back out the same interface through which it was learned.
• Poison Reverse: Explicitly advertising a failed route with an unreachable metric (infinity / 16 hops in RIP) to immediately invalidate neighbor cache entries.
• Holddown Timers: Preventing a router from accepting new advertisements for a flapping route for a designated period, allowing network stability to restore.

Link-State and Dijkstra's Algorithm

Link-state protocols abandon routing by rumor. Instead of advertising forwarding tables, each router discovers its local interfaces and neighbors, encapsulates this information into Link-State Advertisements (LSAs), and floods them to all routers within the area. Every router in that area stores these LSAs in an identical Link-State Database (LSDB).

Once synchronized, each router independently runs the Dijkstra Shortest Path First (SPF) algorithm with itself as the root of the tree, calculating the shortest loop-free path to every destination prefix. Although highly resilient and fast-converging, running SPF across large enterprise topologies consumes substantial CPU and memory, requiring hierarchical area segmentation to bound calculation domains.

---

EIGRP and the Diffusing Update Algorithm (DUAL)

Cisco's Enhanced Interior Gateway Routing Protocol (EIGRP) combines the rapid convergence of link-state protocols with the operational simplicity of distance vector. EIGRP does not maintain a complete topology graph; rather, it uses the Diffusing Update Algorithm (DUAL) to track neighbor reachability and guarantee loop-free backup routes.

EIGRP Composite Metric Formula

Classic EIGRP calculates a 32-bit composite metric using five configurable weighting coefficients ($K_1$ through $K_5$):

$\text{Metric} = 256 \times \left( K_1 \times \text{Bandwidth} + \frac{K_2 \times \text{Bandwidth}}{256 - \text{Load}} + K_3 \times \text{Delay} \right) \times \left( \frac{K_5}{\text{Reliability} + K_4} \right)$

By default, Cisco IOS sets $K_1 = 1$, $K_3 = 1$, and $K_2 = K_4 = K_5 = 0$. When default $K$-values are active, the complex formula reduces to:

$\text{Metric} = 256 \times (\text{Bandwidth} + \text{Delay})$

Where:

• Bandwidth: $\frac{10^7}{\text{Minimum Interface Bandwidth along path (in kbps)}}$
• Delay: $\frac{\text{Cumulative Interface Delay along path (in microseconds)}}{10}$

(Note: Setting $K_5 = 0$ removes the reliability multiplier term from the equation entirely rather than multiplying by zero). Both the EIGRP Autonomous System number and $K$-values must match identically between adjacent routers to form a neighbor relationship.

DUAL Terminology and Loop Freedom

  [Neighbor Router] -------- Advertises Reported Distance (RD) --------> [Local Router]
          |                                                                    |
    Direct Path                                                          Total Metric
    to Destination                                                       to Destination
          |                                                                    |
          v                                                                    v
   [Destination Subnet] <--------------------------------------------- [Feasible Distance (FD)]

• Feasible Distance (FD): The lowest calculated metric from the local router to the destination subnet across all viable paths.
• Reported Distance (RD) / Advertised Distance (AD): The metric to the destination subnet as reported by an adjacent neighbor router (the neighbor's own calculated metric).
• Successor: The primary neighbor path offering the lowest overall metric (FD) to the destination. The successor route is installed into the active routing table.
• Feasible Successor (FS): A backup neighbor that satisfies the Feasibility Condition. An FS is pre-computed and stored in the EIGRP topology table, enabling instantaneous (sub-second) failover if the primary successor fails.
• Feasibility Condition (FC): A neighbor qualifies as a Feasible Successor if and only if:

$\text{Reported Distance (RD)} < \text{Current Feasible Distance (FD)}$

Why the Feasibility Condition Guarantees Loop Freedom: If a neighbor's metric to reach a destination is strictly less than your current best metric (FD), that neighbor cannot possibly be routing packets back through you to reach that destination. If its path looped through you, its reported distance would necessarily be higher than your feasible distance.

Step-by-Step DUAL Calculation Walkthrough

Consider Router R1 evaluating three potential paths to reach destination network 192.168.100.0/24:

Check the arithmetic with the default K-values. For the path through R2:

$$\text{Metric}_{R2} = 256 \times \left(\frac{10^{7}}{100{,}000} + \frac{700}{10}\right) = 256 \times (100 + 70) = 43{,}520$$

For the path through R3, the 10 Mbps link dominates: $256 \times (1{,}000 + 140) = 291{,}840$. R3's reported distance is calculated the same way from R3's own path (100 Mbps minimum bandwidth and 400 $\mu s$ of delay): $256 \times (100 + 40) = 35{,}840$.

If the primary link through R2 fails, R1 immediately promotes R3 as the new successor without recalculating or sending DUAL query packets. Path R4 remains available in the topology table, but if both R2 and R3 fail, R1 must place the route into an Active state and transmit DUAL queries to determine whether R4 offers a loop-free path.

Load Balancing: EIGRP vs. OSPF

Topic 3.2.a also asks you to compare load balancing:

In the walkthrough above, R3's metric is 291,840 and the FD is 43,520, a ratio of about 6.7. Configuring variance 7 under the EIGRP process would install R3 alongside R2. R4 can never be used for unequal-cost load balancing, however high the variance, because it is not a feasible successor.