Interior gateway protocol
An Interior Gateway Protocol (IGP) is a type of routing protocol used within a single autonomous system (AS) to exchange topology and reachability information among routers, allowing them to dynamically build and update routing tables for efficient internal packet forwarding.[1] Unlike Exterior Gateway Protocols (EGPs) such as BGP, which handle inter-AS routing across the internet, IGPs operate solely within the boundaries of one administrative domain to ensure optimal path selection inside enterprise, campus, or service provider networks.[2] IGPs are classified into three main categories based on their operational mechanisms: distance-vector, link-state, and hybrid protocols.[3] Distance-vector protocols, exemplified by Routing Information Protocol (RIP) and Interior Gateway Routing Protocol (IGRP), function by having routers periodically advertise their full routing tables to directly connected neighbors, using metrics like hop count to determine paths; this approach is simple but can lead to slower convergence in larger networks due to potential routing loops.[3] Link-state protocols, such as Open Shortest Path First (OSPF) and Intermediate System to Intermediate System (IS-IS), instead flood link-state advertisements (LSAs) across the AS to construct a complete network topology database on each router, then apply algorithms like Dijkstra's shortest path first (SPF) for rapid, loop-free route calculations, making them suitable for large-scale environments.[3] Hybrid protocols, like Enhanced Interior Gateway Routing Protocol (EIGRP), developed by Cisco and published as an open standard in 2013, blend distance-vector efficiency with link-state features, such as event-driven updates and a topology table for partial network awareness, achieving faster convergence and support for multiple protocols including IP, IPX, and AppleTalk.[4] These protocols play a critical role in modern networking by adapting to changes in topology, such as link failures or additions, through mechanisms like hello packets for neighbor discovery and metrics incorporating bandwidth, delay, or load to prioritize paths.[3] Standardization efforts by bodies like the Internet Engineering Task Force (IETF) have evolved IGPs to support advanced features, including IPv6, traffic engineering, and integration with MPLS for enhanced scalability in data centers and wide-area networks.[5]Introduction
Definition and Purpose
An Interior Gateway Protocol (IGP) is a type of routing protocol designed to exchange routing information between routers within a single autonomous system (AS), enabling the determination of optimal paths for data packets across internal network segments.[6] IGPs operate exclusively within one administrative domain, such as an enterprise network or an ISP's internal infrastructure, without extending to external systems.[7] The primary purpose of an IGP is to facilitate dynamic routing, where routers automatically update their routing tables in response to network changes, providing fault tolerance by rerouting traffic around failures and ensuring efficient packet forwarding.[8] This contrasts with static routing, which requires manual configuration and lacks adaptability to topology alterations.[9] By enabling real-time convergence, IGPs maintain network reliability and performance within the AS boundaries. Key characteristics of IGPs include their operation at the OSI model's network layer (Layer 3), where they process IP addresses and paths, and their use of metrics—such as hop count or bandwidth—to evaluate and select the shortest or most efficient routes.[8] Unlike Exterior Gateway Protocols (EGPs), which handle inter-AS routing, IGPs are confined to intra-AS scope for focused, policy-driven optimization.[7]Relation to Exterior Gateway Protocols
Exterior Gateway Protocols (EGPs) are routing protocols designed to exchange routing information between different autonomous systems (ASes), enabling inter-domain routing across the Internet. Unlike IGPs, which operate within a single AS, EGPs such as the Border Gateway Protocol (BGP) focus on policy-based decisions to determine optimal paths between ASes, considering factors like path attributes, administrative preferences, and scalability for global connectivity.[10][11] The primary differences between IGPs and EGPs lie in their scope, priorities, and operational characteristics. IGPs emphasize rapid convergence and simplicity to maintain efficient internal routing within an AS, often using metrics like hop count or link cost for quick path optimization. In contrast, EGPs prioritize stability and policy enforcement to handle the vast scale of inter-AS routing, resulting in slower convergence but greater resilience against loops and instability across diverse administrative domains.[12][13] IGPs and EGPs interact primarily at the borders of an AS, where internal routes learned via IGPs are redistributed into EGPs to advertise reachability to external networks. This process, known as route redistribution, allows border routers running BGP to inject summarized IGP routes into the EGP, ensuring that external peers can forward traffic into the AS while preventing full internal topology details from being exposed. Synchronization mechanisms, such as next-hop resolution using IGP routes, further ensure that BGP advertisements remain valid based on internal connectivity.[11][14] In the broader hierarchy of Internet routing, IGPs manage intra-domain paths for optimized local traffic flow, while EGPs handle inter-domain exchanges to form the global routing backbone. This layered approach allows each AS to maintain independent internal policies via IGPs, presenting a cohesive external view through EGPs, thereby scaling the Internet's routing infrastructure effectively.[10][15]History and Development
Origins in Early Internet
The origins of Interior Gateway Protocols (IGPs) emerged from the pioneering efforts in packet-switched networking during the 1970s, rooted in the ARPANET project sponsored by the U.S. Department of Defense Advanced Research Projects Agency (DARPA). Launched in 1969, ARPANET connected research institutions with Interface Message Processors (IMPs) that served as foundational routers, implementing dynamic routing to manage topology changes and ensure reliable packet delivery across the network. These IMPs exchanged routing information periodically with neighbors to update paths based on link delays and availability, addressing the limitations of static routing in evolving environments.[16] This approach prioritized adaptability in distributed systems, setting the stage for IGPs to handle internal network updates efficiently.[17] Parallel innovations at Xerox Palo Alto Research Center (PARC) further shaped early IGP designs through the PARC Universal Packet (PUP) protocol suite, developed in the mid-1970s for internetworking heterogeneous networks. PUP employed a simple gateway-based routing mechanism where internetwork gateways propagated distance metrics to direct packets toward destinations, using a hop-count or delay-based evaluation akin to early distance-vector methods. This design influenced the Xerox Network Systems (XNS) architecture in the late 1970s, whose Routing Information Protocol (RIP) variant became a direct precursor to standardized IGPs by enabling automated route exchange among internal nodes.[18][19] The 1980s marked a pivotal shift from predominantly static routing to dynamic protocols, coinciding with the rise of TCP/IP as the foundational internetworking suite. Prior to TCP/IP, networks like ARPANET relied on proprietary dynamic schemes, but the protocol's adoption standardized scalable internal routing needs. On January 1, 1983—known as Flag Day—ARPANET fully transitioned from the Network Control Protocol (NCP) to TCP/IP, amplifying the demand for IGPs to manage route computation within growing autonomous domains.[20] Key IGP concepts were formalized in RFC 791, which defined the Internet Protocol (IP) in September 1981 and outlined datagram routing across interconnected packet-switched networks, termed a "catenet." The specification emphasized gateways for forwarding packets between local networks, requiring internal coordination protocols to exchange topology data and optimize paths within the system; subsequent developments like the Gateway-to-Gateway Protocol (GGP) fulfilled these needs. This underscored the necessity for efficient internal routing to minimize delays and support resource sharing in autonomous networks, distinguishing it from inter-domain concerns.[21]Key Milestones and Standardization
The standardization of interior gateway protocols (IGPs) began in the late 1980s, building on early routing concepts from the ARPANET era. The Routing Information Protocol (RIP) was formally standardized in RFC 1058 in June 1988, defining a distance-vector protocol for exchanging routing information within small to medium-sized networks. This initial specification established RIP as a foundational IGP, relying on hop count as the primary metric. Subsequent enhancements came with RIP version 2 (RIPv2) in RFC 2453 in November 1998, which introduced support for subnet masks, next-hop addresses, and authentication mechanisms like MD5 to address security vulnerabilities in the original version. Parallel developments occurred with link-state protocols. The Open Shortest Path First (OSPF) protocol was first standardized by the IETF in RFC 1131 in October 1989, introducing a robust link-state algorithm for larger, hierarchical networks. OSPF evolved significantly with version 2 (OSPFv2) detailed in RFC 2328 in April 1998, optimizing it for IPv4 environments through improvements in area support, authentication, and scalability features.[22] The Intermediate System to Intermediate System (IS-IS) protocol, originally designed for OSI networks, saw key adaptations for IP use. It was extended for TCP/IP routing in RFC 1195 in December 1990, enabling its deployment as an IGP in IP-dominant environments while maintaining compatibility with OSI. The full ISO standardization of IS-IS followed as ISO/IEC 10589 in July 1992, formalizing its intra-domain routing capabilities.[23] Proprietary advancements also played a major role in IGP evolution during this period. Cisco Systems developed the Interior Gateway Routing Protocol (IGRP) in the mid-1980s as a distance-vector IGP to overcome limitations of RIP, such as using a composite metric including bandwidth and delay for better path selection in larger networks. This was further enhanced with the Enhanced Interior Gateway Routing Protocol (EIGRP) in the early 1990s, introducing hybrid features like partial updates and a dual topology table for faster convergence and support for multiple network layer protocols.[24][25] More recent advancements have focused on IPv6 integration and enhanced security. OSPF version 3 (OSPFv3) was specified in RFC 5340 in July 2008 to support IPv6, incorporating link-local addressing and extending the protocol's flooding and area mechanisms to the new address space. This update also integrated IPsec for authentication and encryption, providing a standardized framework for securing OSPF communications against eavesdropping and spoofing.Classification of IGPs
Distance-Vector Protocols
Distance-vector protocols are a class of interior gateway protocols (IGPs) that enable routers within an autonomous system to exchange routing information by periodically sharing their entire routing tables with directly connected neighbors.[19] This approach relies on the distributed implementation of the Bellman-Ford algorithm, where each router maintains a vector of distances to all known destinations and updates these distances based on information received from neighbors.[19] The metric typically used is a measure of distance, such as hop count, where each link contributes a cost of 1, allowing routers to select paths that minimize this cumulative value.[19] The core update mechanism operates as follows: upon receiving a neighbor's routing table, a router recalculates its own distances using the formula for the minimum path cost: d(x, y) = \min_{z \in N} \left( d(z, y) + c(x, z) \right) where d(x, y) is the distance from router x to destination y, N is the set of neighbors of x, and c(x, z) is the link cost between x and z. For hop count, c(x, z) = 1, so the new distance is simply the neighbor's reported distance plus one; if this yields a shorter path, the router adopts it and sets the neighbor as the next hop.[19] Updates are broadcast periodically, with intervals varying by protocol (e.g., 30 seconds in RIP or 90 seconds in IGRP), ensuring gradual propagation of routing changes across the network.[19][26] These protocols are characterized by their simplicity and periodic nature, which facilitate easy implementation in resource-constrained environments.[19] However, they are susceptible to routing loops, where inconsistent updates can cause temporary cycles, as the distributed Bellman-Ford process may not immediately detect topology changes.[19] To mitigate this, techniques such as split horizon—where routes learned from a neighbor are not advertised back to that same neighbor—and poisoned reverse are employed; the latter advertises invalid routes with a maximum metric (e.g., 16 in RIP for unreachable destinations) to accelerate loop detection and convergence.[19] A key advantage of distance-vector protocols is their low computational overhead, as routers perform only straightforward minimum selections during updates rather than maintaining a full network topology map, making them suitable for smaller networks compared to approaches that flood link-state advertisements.[19] This ease of implementation stems from the minimal state each router needs to track, primarily just the distance vectors per destination.[19]Link-State Protocols
Link-state protocols function as a category of interior gateway protocols in which routers exchange detailed information about their local links to collectively construct a comprehensive map of the network topology. Each router originates link-state advertisements (LSAs) that encapsulate the status and cost of its adjacent links, including connectivity and metrics, and floods these advertisements throughout the routing domain to ensure synchronization.[27] This dissemination process allows every participating router to assemble an identical link-state database, modeling the network as a weighted graph where nodes represent routers and edges denote links.[28] With the topology database in place, each router independently executes the Dijkstra shortest path first (SPF) algorithm on the graph to determine optimal routes to all destinations, guaranteeing consistent path selections across the network. Link costs within this graph are administrator-assigned values, often derived from bandwidth, delay, or a combination thereof, to reflect real-world performance characteristics.[27] A defining feature of link-state protocols is their reliance on event-triggered updates, where LSAs are propagated solely in response to detected topology changes, enabling rapid convergence in contrast to the periodic broadcasts of distance-vector protocols.[29] To address scalability in expansive networks, these protocols incorporate hierarchical structures that confine flooding to defined subdivisions, thereby reducing computational and bandwidth overhead. Their primary advantages stem from this unified topology perspective, which yields precise, loop-free forwarding paths and superior performance in large-scale environments by minimizing inconsistencies and transient errors.[27]Hybrid Protocols
Hybrid protocols combine elements of distance-vector and link-state routing to address limitations of both approaches. They typically use distance-vector messaging for updates but incorporate link-state-like features, such as partial topology maps and event-driven notifications, for improved convergence and efficiency. A prominent example is Enhanced Interior Gateway Routing Protocol (EIGRP), a Cisco proprietary protocol that maintains a topology table and supports multiple network layer protocols.[8]Specific Protocols
Routing Information Protocol (RIP)
The Routing Information Protocol (RIP) is a classic distance-vector interior gateway protocol that employs hop count as its primary routing metric, with a maximum of 15 hops to prevent routing loops and limit its applicability to smaller networks.[30] Developed in the late 1970s as part of Xerox's early networking efforts within the PUP and XNS protocol suites, where it originated as the Gateway Information Protocol (GWINFO), RIP was later standardized by the Internet Engineering Task Force (IETF) for use in IP-based autonomous systems.[30] This metric treats each router hop as a cost of 1, simplifying path selection but ignoring factors like bandwidth or delay, which makes it suitable for straightforward, low-complexity environments.[30] RIP has evolved through two primary versions to address limitations in address handling and security. RIPv1, specified in RFC 1058 (1988), operates as a classful protocol that does not include subnet mask information in updates, assuming network classes based on the first octet of the IP address.[30] In contrast, RIPv2, defined in RFC 2453 (1998), introduces classless routing by incorporating a 32-bit subnet mask field in each route entry, enabling support for variable-length subnet masking (VLSM) and more efficient IP address utilization in diverse network topologies.[31] Additionally, RIPv2 enhances security with optional MD5 authentication (as detailed in RFC 2082), allowing routers to verify the authenticity of incoming updates using a shared key to mitigate spoofing risks. In operation, RIP routers maintain a routing table and periodically exchange information with neighbors to build a network topology view. Every 30 seconds, routers broadcast (in RIPv1) or multicast (in RIPv2, to 224.0.0.9) their full routing tables, enabling incremental updates through the Bellman-Ford algorithm for distance-vector computation.[30] To combat routing loops, RIP implements hold-down timers, which suppress acceptance of worse routes for a route entry for 180 seconds after a failure is detected, alongside techniques like split horizon and poison reverse that prevent advertisements of routes back to their source interface.[30] Invalid routes are marked after 180 seconds of inactivity and flushed after 240 seconds, promoting eventual consistency.[30] Despite its simplicity, RIP exhibits key limitations that restrict its use in modern networks. Convergence can be slow, potentially taking several minutes in cases of topology changes due to the count-to-infinity problem, where routers incrementally increase metrics until reaching the 16-hop infinity threshold.[30] The 15-hop limit inherently caps scalability, making RIP impractical for networks exceeding about 50 routers, as larger topologies would exceed the diameter and render distant destinations unreachable.[30]Open Shortest Path First (OSPF)
Open Shortest Path First (OSPF) is a link-state interior gateway protocol developed by the Internet Engineering Task Force (IETF) and standardized in RFC 2328 for version 2, enabling efficient routing within a single autonomous system by maintaining a consistent topology database across all routers.[32] This database allows each router to independently compute a shortest-path tree to every destination using Dijkstra's algorithm, supporting rapid route recalculation in response to network topology changes.[32] To enhance scalability in large networks, OSPF divides the autonomous system into areas, with Area 0 designated as the backbone area that interconnects all other areas and ensures contiguous routing connectivity.[32] Key components of OSPF include Link State Advertisements (LSAs), which serve as the building blocks of the topology database, and hello packets for neighbor discovery and adjacency maintenance.[32] LSAs are categorized into several types: Type 1 Router-LSAs describe a router's local state, including its links and interfaces; Type 2 Network-LSAs represent multi-access networks and list attached routers; Type 3 Summary-LSAs advertise inter-area network routes; Type 4 Summary-LSAs indicate the location of autonomous system boundary routers; and Type 5 AS-External-LSAs describe routes external to the autonomous system.[32] Hello packets, transmitted periodically to the multicast address 224.0.0.5 on broadcast networks, facilitate neighbor detection, bidirectional communication verification, and the election of designated routers.[32] In operation, OSPF's multi-area architecture confines link-state flooding to individual areas, minimizing protocol overhead and improving convergence compared to flat topologies.[32] On multi-access networks such as broadcast or non-broadcast multiple access (NBMA) segments, OSPF elects a Designated Router (DR) and Backup Designated Router (BDR) to reduce adjacency formation and flooding; the DR originates Type 2 LSAs and relays updates, while the BDR assumes this role upon DR failure.[32] This hierarchical design allows for efficient scaling, with area border routers summarizing routes between areas to prevent unnecessary detail propagation.[32] OSPF employs a cost-based metric for path selection, where the default reference bandwidth is 100 Mbps, enabling administrators to assign costs inversely proportional to link speeds for optimized routing.[32] Additionally, it supports equal-cost multipath (ECMP) routing, permitting load balancing across multiple paths of identical cost, type, and area, provided they have distinct next hops, thereby enhancing network utilization and resilience.[32]Operational Principles
Route Calculation and Metrics
Interior Gateway Protocols (IGPs) perform route calculation by constructing and updating routing tables derived from exchanged protocol messages, enabling routers within an autonomous system to determine optimal paths to destinations. In distance-vector IGPs, such as those based on the Bellman-Ford algorithm, routers propagate distance estimates additively from neighbors, incrementally building knowledge of network distances without a full topology view. Conversely, link-state IGPs, like those utilizing Dijkstra's Shortest Path First (SPF) algorithm, construct a complete graph of the network topology and compute the shortest paths from the local router to all destinations using graph theory principles.[25][5][33] The route calculation process in IGPs follows a structured sequence of steps to ensure consistent and reliable path determination. Neighbor discovery begins with routers sending periodic hello packets or broadcasts to detect and authenticate adjacent devices, establishing adjacency relationships essential for reliable communication. This is followed by information exchange, where distance-vector protocols periodically advertise their entire routing tables to neighbors, while link-state protocols flood detailed link-state updates across the network to synchronize topology databases among all routers. Path selection then applies the protocol's algorithm to evaluate exchanged data against configured metrics, identifying the lowest-cost route to each destination. Finally, the computed routes are installed into the IP routing table, updating the forwarding information base (FIB) to guide packet forwarding decisions.[33][25] Metrics serve as the quantitative criteria for path evaluation in IGPs, quantifying the desirability of routes based on network characteristics to prefer efficient paths. A basic metric is hop count, which simply tallies the number of intermediate routers to a destination, offering simplicity but failing to account for variations in link bandwidth or latency, potentially leading to suboptimal traffic flows on congested or slow links. More advanced composite metrics integrate multiple factors, such as bandwidth (prioritizing high-capacity links), delay (minimizing propagation time), load (avoiding overloaded interfaces), and reliability (favoring stable connections), to provide a holistic assessment of path quality; for instance, bandwidth and delay are often weighted together to balance throughput and responsiveness.[3][34][25] To enhance network utilization, IGPs implement load balancing through Equal-Cost Multi-Path (ECMP) when multiple candidate paths exhibit identical metrics, allowing traffic to be distributed across them via hashing or round-robin methods, thereby improving throughput and resilience without compromising path optimality. In protocols like OSPF, this calculation occurs within defined areas to hierarchically manage complexity and scale.[35][33]Convergence Mechanisms
Convergence in interior gateway protocols (IGPs) refers to the process by which all routers in a network reach a consistent, loop-free view of the topology following a change, such as a link failure or addition. This agreement ensures that routing tables are updated to reflect the new shortest paths without temporary loops or blackholes, minimizing packet loss and downtime. In practice, convergence time is measured from the detection of a topology event until all routers forward traffic along the updated paths, and it varies significantly between protocol types due to their underlying mechanisms.[36] Link-state protocols, such as OSPF, achieve faster convergence—typically within seconds—through triggered updates that immediately flood link-state advertisements (LSAs) across the network upon detecting a change. Routers then perform a shortest path first (SPF) recalculation using Dijkstra's algorithm on their synchronized link-state databases, ensuring a loop-free topology is established quickly without propagating outdated information hop-by-hop. This process is supported by timers like the Hello interval (default 10 seconds on broadcast networks) for neighbor detection and the Dead interval (default 40 seconds) to identify failed links, with reliable flooding and acknowledgments preventing loss of updates. In small to medium networks, this results in near-instantaneous convergence, as the flooding propagates changes in O(E) time where E is the number of edges, followed by efficient SPF computation.[37][38][39] Distance-vector protocols, like RIP, rely on periodic updates every 30 seconds to exchange routing tables with neighbors, leading to slower convergence that can take minutes in larger networks. Triggered updates are sent for route improvements or upon detecting increases (e.g., via route poisoning with metric 16 for unreachable destinations), but bad news about failures propagates gradually, potentially up to O(N) time where N is the network diameter in hops, due to mechanisms like split horizon and hold-down timers that suppress invalid routes for 180 seconds to prevent loops. Additional techniques include invalid timers (180 seconds to mark routes invalid) and garbage collection (120 seconds to fully remove them), which help stabilize the network but delay full agreement on the topology. Route damping, implemented via these hold-down and suppression timers, mitigates flapping routes by temporarily ignoring unstable updates to avoid repeated reconvergences.[40][41] Overall, while link-state IGPs prioritize speed through global topology awareness, distance-vector approaches balance simplicity with stability features, though at the cost of longer convergence in expansive topologies. Route metrics, such as hop count in RIP or composite costs in OSPF, indirectly influence convergence by affecting how quickly alternative paths are selected during recalculation.[42]Applications and Considerations
Use in Autonomous Systems
Interior Gateway Protocols (IGPs) are deployed in various contexts within autonomous systems (ASes) to manage internal routing efficiently. In enterprise local area networks (LANs), particularly smaller setups, the Routing Information Protocol (RIP) is commonly used due to its simplicity and suitability for networks with limited routers, where hop count serves as the primary metric for path selection.[43][44] For larger enterprise environments, RIP's limitations in scalability lead to preferences for more advanced protocols, but it remains a foundational choice for basic connectivity in corporate LANs. In service provider core networks, link-state protocols such as Open Shortest Path First (OSPF) and Intermediate System to Intermediate System (IS-IS) dominate deployments, handling high-scale routing across extensive topologies. OSPF is often configured with a hierarchical area structure to support connectivity between core devices, enabling rapid path computations via the Dijkstra algorithm.[45] IS-IS is particularly favored in provider backbones for its flexibility in large, flat designs without rigid area dependencies, allowing seamless scaling in multi-protocol environments. These protocols ensure low-latency forwarding in carrier-grade infrastructures by flooding link-state advertisements (LSAs) to maintain a synchronized topology view.[46] Data centers leverage fast-converging link-state IGPs like OSPF and IS-IS to support dynamic workloads and minimize downtime from failures. These protocols achieve sub-second convergence times through optimized shortest path first (SPF) calculations, which is critical for high-availability fabrics where traffic demands rapid rerouting.[47] In spine-leaf architectures, OSPF's multi-area support or IS-IS's level-based hierarchy facilitates efficient underlay routing, ensuring low overhead in environments with thousands of endpoints.[48] IGPs integrate with Multiprotocol Label Switching (MPLS) to enable traffic engineering, allowing explicit path control beyond default IGP shortest paths. In MPLS networks, IGPs such as OSPF or IS-IS distribute link metrics and attributes via extensions like opaque LSAs, while protocols like Resource Reservation Protocol-Traffic Engineering (RSVP-TE) establish label-switched paths (LSPs) for optimized bandwidth utilization.[49][50] This synergy supports constrained shortest path first (CSPF) computations, directing traffic over engineered routes to balance loads and avoid congestion in backbone networks.[51] Hybrid integration with Software-Defined Networking (SDN) combines IGP reliability with centralized control, using IGPs as underlays for distributed forwarding while SDN controllers handle overlay orchestration. In such setups, protocols like OSPF provide baseline topology awareness to SDN applications, enabling automated policy enforcement and resilience through failover mechanisms.[52] This approach is common in evolving networks transitioning to full SDN, where IGPs ensure robust connectivity during phased deployments.[53] Best practices for OSPF deployment emphasize thoughtful area design to minimize LSA flooding and enhance scalability. Dividing an AS into multiple areas, with a backbone area (Area 0) connecting all others, confines LSAs to their originating areas, reducing the link-state database size on individual routers and accelerating SPF runs.[54] Limiting area border routers (ABRs) to 2–4 per area further curbs summary LSAs, preventing overload in stub or totally stubby areas that block external route advertisements.[55] Metric tuning in OSPF facilitates load balancing by adjusting interface costs to influence path selection. The default cost, inversely proportional to bandwidth (e.g., reference bandwidth of 100 Mbps divided by interface speed), can be manually set via theip ospf cost command to equalize paths, allowing up to four equal-cost routes in the routing table for per-packet or per-destination distribution.[56] This technique promotes even utilization of parallel links, such as in multi-homed access layers, without requiring unequal-cost forwarding extensions.[57]
In BGP confederations, IGPs play a key role by managing routing within sub-ASes, thereby reducing the need for a full iBGP mesh across the entire AS. Each sub-AS runs an IGP like OSPF or IS-IS for internal reachability, while confederation eBGP handles inter-sub-AS peering, limiting iBGP sessions to within smaller domains and simplifying scale in large networks.[58][59] This design interacts with exterior gateway protocols at AS borders through route redistribution, ensuring seamless external connectivity.