When a packet needs to travel from one end of an enterprise campus to another, or from a data center edge to a core router within the same organization, it doesn't need to understand the complexities of global Internet routing. It doesn't need to know about AS paths, routing policies, or inter-organizational agreements. It simply needs to find the best path within its local domain.

This is the realm of Interior Gateway Protocols (IGPs)—the routing protocols designed to efficiently and automatically determine optimal paths within a single Autonomous System. IGPs are the workhorses of enterprise, data center, and service provider networks, running continuously to ensure that internal connectivity remains optimal even as conditions change.

Interior Gateway Protocols share a common purpose: determine the best paths for traffic within a single administrative domain. But what exactly distinguishes an IGP from its exterior counterparts?

Fundamental IGP Characteristics:

IGP vs. EGP Philosophy:

The fundamental difference between IGP and EGP is philosophy:

Within your AS, you want the fastest, most efficient paths. Between ASes, you want paths that comply with business agreements, legal requirements, and security policies—even if they're not technically 'optimal.'

All IGPs are built on one of two fundamental routing algorithms: Distance Vector or Link State. Understanding these algorithms is essential for understanding how different IGPs behave, their strengths, their limitations, and why network architects choose one over another.

Note: EIGRP is sometimes called an "advanced distance vector" or "hybrid" protocol because it incorporates some link-state features like triggered updates and maintaining neighbor relationships, while fundamentally using distance vector principles.

The Routing Information Protocol (RIP) is one of the oldest routing protocols still in use. First standardized in RFC 1058 (1988), RIP traces its origins to the Xerox PARC network in the 1970s and BSD Unix's routed daemon. While largely superseded in enterprise networks, RIP remains educational as the canonical distance vector implementation.

RIP Versions:

RIP Operation:

1. Metric: Hop count (number of routers to destination). Maximum is 15; 16 means "unreachable."

2. Updates: Every 30 seconds, routers broadcast/multicast their complete routing table.

3. Timers:

   • Update timer (30s): Regular full-table advertisements
   • Invalid timer (180s): Route marked invalid if no update received
   • Flush timer (240s): Route removed from table entirely
   • Holddown timer (180s): Prevents accepting inferior routes during convergence

4. Loop Prevention Mechanisms:

   • Split Horizon: Don't advertise a route back to the router you learned it from
   • Split Horizon with Poison Reverse: Advertise learned routes back with metric 16 (unreachable)
   • Holddown: After learning a route is down, ignore updates for a period
   • Triggered Updates: Send immediate updates when routes change (don't wait 30s)

Open Shortest Path First (OSPF) is the most widely deployed IGP in enterprise and service provider networks. Standardized in RFC 2328 (OSPFv2 for IPv4) and RFC 5340 (OSPFv3 for IPv6), OSPF provides:

• Fast convergence through event-driven updates
• Hierarchical network design through areas
• Support for VLSM and CIDR
• Rich metric system based on interface cost
• Extensive authentication options
• Multi-vendor interoperability (open standard)

OSPF Fundamentals:

1. Neighbor Discovery and Adjacency Formation:

• OSPF routers discover neighbors through Hello packets (multicast to 224.0.0.5)
• Neighbors must agree on key parameters: area ID, hello/dead intervals, authentication, stub flags
• Full adjacencies are formed only when needed (based on network type and DR/BDR election)

2. Link State Database (LSDB):

• Every router maintains a synchronized LSDB containing Link State Advertisements (LSAs)
• LSAs describe routers, networks, and their connections
• Flooding ensures all routers in an area have identical LSDB

3. SPF Calculation:

• Each router runs Dijkstra's algorithm on its LSDB
• Results in a shortest-path tree rooted at the computing router
• Routing table is derived from this tree

OSPF Hierarchical Design:

OSPF introduces the concept of areas to enable hierarchical network design:

• Area 0 (Backbone): All other areas must connect to Area 0, either directly or through virtual links
• Regular Areas: Contain internal routers and connect to the backbone
• Stub Areas: Don't receive external (Type 5) LSAs; use a default route instead
• Totally Stubby Areas: Don't receive external or inter-area (Type 3) LSAs
• Not-So-Stubby Areas (NSSA): Allow limited external route injection into stub areas

Areas provide:

• Reduced LSDB size: Routers only need full topology for their own area
• Faster SPF calculation: Smaller database = faster computation
• Contained flooding: Topology changes flood only within an area
• Route summarization: Area Border Routers (ABRs) can summarize routes at boundaries

Intermediate System to Intermediate System (IS-IS) is a link state IGP that, while less common in enterprises, dominates major service provider and hyperscale networks. Originally designed for the OSI protocol stack (ISO 10589), IS-IS was extended to support IP ("Integrated IS-IS") and later IPv6.

Why Service Providers Prefer IS-IS:

IS-IS Fundamentals:

Addressing: IS-IS uses Network Entity Titles (NETs) to identify routers:

• Format: 49.0001.0000.0000.0001.00
• Area ID: 49.0001 (variable length)
• System ID: 0000.0000.0001 (6 bytes, like MAC address format)
• Selector: 00 (always 00 for routers)

Hierarchy:

• Level 1 (L1): Intra-area routing (like OSPF regular area)
• Level 2 (L2): Inter-area routing (like OSPF backbone)
• Level 1-2 (L1/L2): Router that participates in both (like OSPF ABR)

Unlike OSPF, there's no requirement for a contiguous backbone. Any chain of L2-capable routers forms the backbone.

PDU Types:

• IIH (IS-IS Hello): Neighbor discovery and maintenance
• LSP (Link State PDU): Carry topology information
• CSNP (Complete Sequence Number PDU): Database synchronization
• PSNP (Partial Sequence Number PDU): Request specific LSPs or acknowledge receipt

Enhanced Interior Gateway Routing Protocol (EIGRP) is Cisco's proprietary (now partially open-standard) routing protocol that combines distance vector simplicity with link state-like features. Often called a "hybrid" or "advanced distance vector" protocol, EIGRP offers:

• Very fast convergence (often sub-second)
• Loop-free paths guaranteed by DUAL algorithm
• Reduced bandwidth usage (only incremental updates)
• Support for multiple routed protocols (IP, IPv6, legacy IPX/AppleTalk)
• Unequal-cost load balancing capability

The DUAL Algorithm:

EIGRP's Diffusing Update Algorithm (DUAL) is what sets it apart from traditional distance vector protocols:

1. Feasibility Condition: A route is "feasible" (loop-free) if the neighbor's reported distance to a destination is less than the router's current feasible distance. This mathematical guarantee prevents loops without waiting for holddown timers.

2. Successor and Feasible Successor:

   • The Successor is the best route to a destination
   • Feasible Successors are backup routes that meet the feasibility condition
   • When a Successor fails, EIGRP can instantly switch to a Feasible Successor (no SPF calculation needed)

3. Active vs. Passive State:

   • Passive: Route is stable; Successor is known
   • Active: No Feasible Successor available; EIGRP queries neighbors to find alternative paths

The DUAL algorithm provides mathematical correctness guarantees that traditional distance vector protocols lack, while avoiding the computational overhead of full SPF recalculation.

EIGRP Metric Calculation (Simplified):

Metric = 256 × (K1×Bandwidth + K2×Bandwidth/(256-Load) + K3×Delay)
         × K5/(Reliability + K4)

With default K-values (K1=1, K2=0, K3=1, K4=0, K5=0), this simplifies to:

Metric = 256 × (Bandwidth + Delay)

Where:

• Bandwidth = 10^7 / (minimum bandwidth in Kbps)
• Delay = sum of delays in tens of microseconds

Wide Metrics (RFC 7868): Modern EIGRP uses 64-bit wide metrics to handle high-speed interfaces properly, avoiding the issue where all high-speed links calculate identical metrics.

Selecting an IGP is one of the most important network design decisions. The right choice depends on network size, vendor environment, required features, and operational expertise. Here's a decision framework:

Key Decision Factors:

1. Network Scale:

• Small (< 50 routers): Any IGP works; choose based on familiarity
• Medium (50-500 routers): OSPF or EIGRP with proper area/summarization design
• Large (500+ routers): IS-IS or well-designed OSPF with route reflectors/confederation

2. Vendor Mix:

• Cisco-only: EIGRP is a strong option
• Multi-vendor: OSPF or IS-IS (both are open standards)
• Open-source/Linux routers: OSPF (FRRouting, Quagga support excellent)

3. Existing Expertise:

• Don't underestimate operational familiarity
• Migrating protocols is costly; choose something the team knows or can learn

4. Future Requirements:

• MPLS Traffic Engineering: IS-IS has best TE integration
• Segment Routing: Both OSPF and IS-IS support SR extensions
• IPv6: OSPFv3 or IS-IS with IPv6 AF

5. Convergence Requirements:

• Subsecond: EIGRP (with feasible successors) or OSPF/IS-IS with BFD
• Standard: Any IGP with tuned timers
• Best-effort: Default configurations work

We've explored the rich landscape of Interior Gateway Protocols. Let's consolidate the essential knowledge:

What's Next:

Having mastered IGP protocols, we'll now turn to EGP protocols—specifically BGP, the protocol that glues the Internet together. Understanding how traffic moves between Autonomous Systems is essential for anyone working with Internet-connected infrastructure.