Fact-checked by Grok 2 weeks ago

Internet transit

Internet transit is a commercial service offered by Internet Service Providers (ISPs), known as transit providers, that enables customers—such as smaller ISPs, content providers, or enterprises—to gain full access to the global Internet by routing their outbound traffic through the provider's network and receiving inbound traffic from the rest of the Internet in return.^[1] This bidirectional connectivity is achieved through the exchange of routing information via protocols like the Border Gateway Protocol (BGP), allowing the transit provider to announce customer routes to the broader Internet while advertising routes to all publicly reachable destinations back to the customer.^[2] Unlike direct connections, transit relies on the provider's infrastructure, including high-capacity backbone networks and interconnections with other autonomous systems (AS), to ensure seamless data flow across diverse geographies.^[3] The service operates on a customer-supplier model, where customers pay the transit provider based on bandwidth usage, typically measured using the 95th percentile method to bill for peak traffic without penalizing short bursts.^[1] Contracts often include service level agreements (SLAs) guaranteeing performance metrics like latency, packet loss, and uptime, with port speeds ranging from gigabit Ethernet to 400 gigabit Ethernet or higher.^[2] Transit providers, particularly Tier 1 networks, connect directly to each other without upstream dependencies, forming the Internet's core backbone with immense capacities—such as over 550 terabits per second in some cases—enabling global scalability.^[2] A key distinction from Internet peering is that transit is a paid arrangement providing unrestricted access to the entire Internet, including routes from upstream providers and peers, whereas peering involves settlement-free exchanges of traffic between networks of comparable size for mutual benefit, limited to each other's customers.^[3] This makes transit essential for networks lacking extensive direct interconnections, ensuring universal reachability but at a cost that has historically declined by approximately 30% annually since the late 1990s, driven by traffic growth exceeding 50% per year and infrastructure efficiencies.^[1] Economically, Internet transit underpins the Internet's hierarchical structure, supporting everything from residential broadband to cloud services by aggregating traffic and optimizing paths through BGP route announcements.^[2] Its importance lies in democratizing connectivity, allowing even small entities to participate in the global ecosystem, though reliance on fewer transit providers can introduce risks like single points of failure or geopolitical dependencies.^[3]

Fundamentals

Definition

Internet transit is a business and technical service in which an upstream Internet service provider (ISP), known as a transit provider, enables a customer network—such as a downstream ISP or enterprise—to exchange traffic with the entire Internet.^[4] This service provides paid access to global reachability, allowing the customer to send and receive data from any Internet destination without needing direct connections to every other network.^[3] Unlike peering, which involves settlement-free traffic exchange limited to each other's customers, transit ensures comprehensive connectivity through the provider's broader network relationships.^[5] For inbound traffic destined to the customer, the transit provider advertises the customer's IP prefixes to its peers and upstream providers using the Border Gateway Protocol (BGP), which enables route advertisement across autonomous systems.^[4] This advertisement informs the global Internet routing system of the paths to reach the customer's networks, directing incoming packets through the transit provider's infrastructure to the customer.^[6] For outbound traffic from the customer to the rest of the Internet, the transit provider supplies the customer with either the full Internet routing table or a default route via BGP, allowing the customer's routers to forward packets toward any destination.^[4] This ensures efficient path selection for the customer's traffic without requiring the customer to maintain complete visibility into all Internet routes.^[6] A representative example is a regional ISP purchasing transit from a Tier 1 provider such as AT&T, which leverages its global backbone to connect the regional network to international destinations unreachable through local connections alone.^[7]

Role in Internet Architecture

Internet transit serves as a foundational element in the Internet's ecosystem, acting as the primary mechanism for non-Tier 1 networks to achieve global reachability. It enables a hierarchical connectivity model where smaller autonomous systems (ASes) purchase transit services from larger providers, forming a structured dependency that underpins the Internet's operation without requiring universal direct interconnections.^[4] This positioning allows transit providers, particularly Tier 1 networks, to function as the core backbone, distributing traffic efficiently across the global topology.^[8] Central to this architecture is the dependency on Autonomous Systems, identified by unique AS numbers assigned by regional Internet registries. Transit arrangements create a dependency tree in the AS graph, where non-Tier 1 ASes rely on upstream providers for routing information and connectivity, culminating in Tier 1 networks at the apex that maintain default-free routing tables—complete global routing information without default routes.^[9] This structure ensures that lower-tier networks can propagate routes upward through providers, propagating reachability downward in a scalable manner. Approximately 80,000 active ASes worldwide as of November 2025, the vast majority of which are non-Tier 1, depend on transit to obtain full Internet access rather than establishing peering relationships with every other network.^[10] By facilitating the exchange of packets across disparate AS boundaries, Internet transit contributes significantly to the end-to-end connectivity principle that defines the Internet's decentralized design. It allows data to traverse multiple independent networks via BGP-advertised routes, ensuring reliable delivery from source to destination without the need for end hosts to manage inter-network paths directly.^[11] This intermediary role supports the Internet's interconnected yet autonomous nature, where transit providers bridge gaps between peering ecosystems.^[12] The use of transit enhances the Internet's scalability by permitting modular network growth; new ASes or expansions can connect via a single or few transit providers to reach the entire Internet, obviating the need for bilateral agreements with thousands of other entities. This modularity has enabled the Internet to accommodate exponential growth in users and traffic while maintaining operational efficiency.^[9]

Historical Development

Origins (1970s–1980s)

The origins of Internet transit trace back to the development of ARPANET, the pioneering packet-switched network funded by the U.S. Department of Defense's Advanced Research Projects Agency (ARPA). Launched in 1969, ARPANET connected four initial nodes at UCLA, Stanford Research Institute, the University of California Santa Barbara, and the University of Utah, enabling researchers to share resources through decentralized data transmission. In this early setup, all participating nodes operated under implicit transit-like assumptions, where each interface message processor (IMP) forwarded packets on behalf of others across the network without any formal payment structures or commercial incentives, relying instead on collaborative resource sharing among academic and military institutions.^[13] This mutual routing model, facilitated by the Network Control Program (NCP), established the foundational principle of interconnected networks providing passage for external traffic as a public good.^[13] A pivotal advancement occurred on January 1, 1983—known as "Flag Day"—when ARPANET fully transitioned to the TCP/IP protocol suite, replacing NCP and standardizing internetworking protocols across the network.^[14] Developed by Vint Cerf and Bob Kahn, TCP/IP enabled more robust, scalable routing by separating transport and network layers, allowing packets to traverse diverse underlying networks with consistent addressing and forwarding mechanisms.^[15] This adoption not only unified ARPANET's operations but also laid the groundwork for hierarchical transit structures by introducing internet protocol (IP) addressing that supported expansion beyond a single network, facilitating interconnections with emerging systems without proprietary barriers.^[16] Throughout the 1980s, the proliferation of academic and research networks amplified these non-commercial transit dynamics, driven by cooperative agreements among institutions. Networks like CSNET (launched in 1981) and BITNET extended connectivity to computer science departments and universities, sharing ARPANET infrastructure through informal arrangements that permitted traffic transit without monetary exchange.^[13] The deployment of NSFNET in 1985 marked a significant escalation, as the National Science Foundation established a 56 kbps backbone linking six supercomputer centers and integrating regional networks such as those operated by mid-level providers.^[17] These regional networks implicitly transited traffic via NSFNET's backbone connections, assuming a full mesh of cooperative access that prioritized research collaboration over profit.^[18] This era culminated in the emergence of a rudimentary hierarchy distinguishing core backbone networks from edge regional ones, with NSFNET serving as a de facto transit provider. By interconnecting supercomputer sites, ARPANET gateways, and over a dozen regional networks, NSFNET's architecture positioned the backbone as the central conduit for inter-regional traffic, enforcing policies that ensured equitable, non-discriminatory forwarding among connected entities.^[19] This core-edge model, supported by the DARPA Internet protocol suite, fostered a scalable framework where peripheral networks relied on the backbone for broader reach, all under grant-funded cooperative governance rather than market-driven contracts.^[20]

Commercialization and Growth (1990s–2000s)

The privatization of the NSFNET backbone in 1995 marked a pivotal shift from government-funded infrastructure to commercial Internet transit services. On April 30, 1995, the National Science Foundation decommissioned the NSFNET, ending its role as the primary U.S. Internet backbone and prompting the emergence of private providers to handle inter-network traffic.^[21] This transition created opportunities for commercial entities like UUNET and PSINet (PSI), which rapidly expanded their backbone networks to connect regional ISPs and enable global reach, filling the void left by the federal network.^[22] These early providers offered paid transit services, allowing ISPs to purchase bandwidth for routing traffic across the Internet, thus laying the foundation for a market-driven ecosystem.^[23] To facilitate interconnections among these emerging commercial networks, the NSF introduced Network Access Points (NAPs) in the early 1990s as neutral hubs for traffic exchange. Contracts for the NAPs were awarded in 1994 following a 1993 solicitation, with the first operational sites—such as MAE-East in Washington, D.C., and others in New York, Chicago, and San Jose—coming online by 1994 and 1995, initially funded by the NSF to bridge the gap during privatization.^[24] These NAPs served as formal transit points where ISPs could connect without direct bilateral agreements, evolving from NSF oversight into commercially operated exchanges that supported the growing volume of inter-provider traffic.^[23] By standardizing interconnection, the NAPs accelerated the commercialization process, enabling scalable transit arrangements under the Border Gateway Protocol (BGP) for route advertisement.^[21] A key milestone occurred in 1998, when the NSF fully withdrew its funding and oversight, transferring NAP operations and related functions like the very high-speed Backbone Network Service (vBNS) to private entities, compelling ISPs to procure transit services exclusively from commercial providers for comprehensive Internet access.^[21] This deadline solidified the transit model, as smaller ISPs relied on Tier 2 providers purchasing from emerging Tier 1 backbones to achieve end-to-end connectivity. The 2000s saw explosive growth in Internet transit driven by the dot-com boom, which surged demand for bandwidth amid e-commerce expansion and web adoption, with global Internet traffic increasing from about 3 Gbps in 1996 to around 21 Tbps by 2000.^[25] This era prompted consolidation among providers; for instance, Level 3 Communications acquired Genuity in 2002 to bolster its IP network capabilities and later integrated WilTel's assets in 2005, enhancing its position as a major transit supplier.^[26] Similarly, WorldCom's earlier acquisitions of UUNET in 1998 and MCI positioned it as a dominant player before its 2002 bankruptcy led to further restructuring.^[27] By 2005, the transit market had evolved from a niche service into a multi-billion-dollar industry, dominated by a handful of Tier 1 providers like AT&T, Level 3, and Sprint that interconnected without purchasing transit from each other.^[28]

Technical Implementation

Routing and BGP Protocols

The Border Gateway Protocol version 4 (BGP-4) serves as the primary inter-domain routing protocol for exchanging reachability information between autonomous systems (ASes) on the Internet. Standardized in RFC 1771 and later updated in RFC 4271, BGP-4 enables ASes to advertise IP prefixes and associated path information, allowing routers to construct paths for forwarding traffic across diverse networks. Unlike interior gateway protocols, BGP operates as a path-vector protocol, maintaining a table of reachable destinations with attributes that influence route selection and policy enforcement.^[29]^[30] In Internet transit arrangements, BGP facilitates the exchange of routing information between a customer AS and its transit provider. The customer announces its own IP prefixes to the provider via BGP updates, enabling the provider to route traffic destined for the customer's networks. In response, the transit provider advertises either its full BGP routing table—comprising over 1 million IPv4 prefixes as of 2025—or a default route to the customer, granting access to the broader Internet. This asymmetric exchange ensures the customer can reach global destinations while the provider controls outbound traffic engineering through selective announcements.^[31]^[32] BGP employs several path attributes to manage route propagation and selection, particularly in transit scenarios. The AS_PATH attribute records the sequence of ASes traversed by a route, prepending the originating AS number to prevent routing loops by discarding paths that include the local AS. LOCAL_PREF, a well-known discretionary attribute, is used internally within an AS to prioritize routes toward preferred transit providers, assigning higher values to customer-learned paths over those from peers or upstreams. The Multi-Exit Discriminator (MED) attribute, optional and non-transitive, allows providers to influence inbound traffic from customers by suggesting preferred entry points into their network, typically set lower for desired paths. These attributes enable fine-grained policy control without altering the core reachability data.^[29]^[33]^[34] BGP sessions are established over TCP connections on port 179 to ensure reliable message delivery between peering routers. During session initialization, routers exchange OPEN messages containing parameters like the hold time, after which periodic KEEPALIVE messages—sent at intervals typically one-third of the hold time (default 180 seconds)—maintain the session's viability. If no KEEPALIVE or UPDATE message is received within the hold timer, the session is torn down to detect failures promptly. This mechanism supports the stability required for exchanging large volumes of routing data in transit environments.^[29]^[35] To enforce contractual boundaries, transit providers implement route filtering policies that prevent customers from inadvertently or maliciously becoming transit ASes for third parties. These policies, often tied to no-transit clauses in service agreements, use access control lists, prefix lists, or AS-path filters to block advertisements of routes learned from one customer to another or to upstream providers. For instance, a provider might filter out routes with customer ASes in the AS_PATH unless explicitly permitted, ensuring traffic flows only through authorized paths and mitigating risks like route leaks. Such filtering is a standard best practice for maintaining network integrity and scalability.^[36]^[37]

Network Configurations and Agreements

Internet transit agreements vary in scope to accommodate different customer requirements for routing information and redundancy. Full route options provide customers with the complete global BGP routing table, approximately 1,038,000 IPv4 prefixes as of November 2025, allowing for fine-grained path control and optimal traffic engineering.^[31] In contrast, partial route agreements deliver a subset of routes, typically including the provider's customer prefixes plus major upstream networks, reducing resource demands on customer routers while still enabling selective routing. Default route setups offer the simplest configuration, supplying only a single default gateway (0.0.0.0/0) for all non-local traffic, suitable for smaller networks with limited BGP capabilities. Multi-homing, where customers connect to multiple transit providers simultaneously, is commonly implemented across these agreement types to ensure redundancy, failover, and improved resilience against single-provider outages.^[4] Physical and logical connections for transit services are established through dedicated infrastructure to ensure reliable, low-latency access. Physical links often utilize fiber optic cross-connects within colocation data centers, where customer equipment is directly patched to the provider's point-of-presence (PoP) via short cable runs, minimizing distance and signal degradation.^[38] Virtual circuits, implemented over Layer 2 Ethernet or MPLS networks, provide logical separation and scalability without requiring additional physical cabling, allowing multiple virtual connections over a single physical port.^[39] Colocation at shared facilities is a standard practice, enabling customers to house routers adjacent to provider gear for sub-millisecond latencies and simplified maintenance.^[38] Service level agreements (SLAs) in internet transit emphasize best-effort delivery for data packets, meaning no absolute guarantees on bandwidth or packet loss beyond committed port speeds, but with defined targets for availability and performance. Uptime guarantees typically range from 99.95% to 99.99% monthly, excluding scheduled maintenance, with financial penalties such as service credits applied for breaches exceeding thresholds like four hours of downtime.^[40] Latency targets, often measured end-to-end, aim for under 50 milliseconds within a region, though these are monitored rather than strictly enforced due to the shared nature of internet routing.^[41] Bandwidth is provisioned as burstable, allowing temporary spikes up to line rate without overage charges, but sustained overuse may trigger throttling to prevent network congestion.^[41] A typical configuration involves the customer establishing a BGP peering session with the transit provider over the dedicated connection, announcing their IP prefixes for inbound routing. Providers enforce prefix limits, such as a minimum /24 for IPv4 announcements (256 addresses), to maintain routing table efficiency and prevent fragmentation. Maximum prefix counts per session, often capped at 100-500, protect against route leaks or attacks, with automatic session teardown if exceeded.^[42] To prevent abuse, agreements include traffic ratio clauses monitoring inbound-to-outbound volumes, typically allowing up to 10:1 inbound without additional fees, beyond which the provider may impose limits or require upgrades to avoid subsidizing excessive peering-like traffic. No-transit clauses in agreements prohibit customers from using the provider's network to carry traffic between third parties, enforced technically via BGP communities. These extended attributes, such as the standard NO_EXPORT community (value 65535:65281), tag routes to restrict propagation beyond the immediate AS, ensuring the provider's infrastructure remains a conduit only for customer-to-internet flows rather than unintended transit paths.^[36] Providers configure outbound policies to append these communities to customer-announced routes advertised to peers, thereby controlling global visibility and upholding contractual boundaries.^[43]

Business Models

Provider Tiers and Relationships

Internet transit providers are classified into a tier system based on their network scope, peering arrangements, and dependency on paid transit services for global reachability. Tier 1 providers operate expansive, global networks that interconnect directly with each other through settlement-free peering agreements, eliminating the need to purchase transit from any other entity to access the entire internet. Examples include Verizon, AT&T, NTT Communications, Lumen Technologies (formerly Level 3), and Tata Communications, which maintain presence across multiple continents and handle intercontinental traffic exchange. These providers form the core backbone of the internet, ensuring default-free routing without relying on external gateways.^[44] Tier 2 providers possess large regional or national networks but purchase transit from one or more Tier 1 providers to achieve full global connectivity, while simultaneously selling transit services to lower-tier customers and engaging in selective peering with other Tier 2 networks.^[45] In contrast, Tier 3 providers, often local or access-focused networks such as regional ISPs or "eyeball" networks serving end-users, exclusively buy transit from higher tiers and do not sell it, focusing instead on last-mile delivery to consumers.^[4] This tiered structure creates hierarchical dependencies, where Tier 2 and Tier 3 networks rely on Tier 1 providers for access to the default-free zone (DFZ)—the comprehensive set of global routes maintained collectively by Tier 1 backbones—resulting in a pyramid-like ecosystem of upstream providers and downstream customers.^[46] Customers of transit providers span various types, including other ISPs seeking broader connectivity, content providers like streaming services requiring reliable upstream paths for data distribution, and enterprises needing dedicated bandwidth for operations; in these relationships, the transit provider acts as the upstream entity supplying reachability, while the customer is downstream, paying for the service to extend their network's scope.^[45] As of 2023, approximately 11 to 16 true Tier 1 providers exist worldwide.^[44] Over time, provider relationships have evolved from predominantly bilateral transit agreements toward more multilateral interconnections, facilitated by internet exchange points (IXPs) that enable efficient traffic exchange among multiple parties; however, paid transit remains essential for lower-tier providers to guarantee complete global reachability beyond localized peering.^[47]

Pricing and Billing Practices

Internet transit services are typically priced on a per-megabit-per-second (Mbps) or per-gigabit-per-second (Gbps) basis per month, with rates varying significantly by region, volume commitment, and provider tier. In competitive markets during 2023, wholesale IP transit prices for 100 Gbps ports ranged from approximately $0.06 per Mbps per month in Europe to $0.35 per Mbps per month in Latin America for 10 Gbps ports, reflecting differences in infrastructure density and demand.^[48]^[49] Customers often commit to a minimum bandwidth level to secure these rates, which can include options for burstable or dedicated capacity. Prices continued to decline into 2025, with European 100 GigE rates around $0.05 per Mbps per month.^[50] A common billing method for burstable transit is the 95th percentile approach, which calculates charges based on peak usage while excluding the highest 5% of traffic bursts to account for occasional spikes rather than sustained demand. Under this model, bandwidth usage is sampled at regular intervals (typically every 5 minutes) over a month, and the 95th percentile value—representing the threshold below which 95% of samples fall—determines the billable rate, promoting fair pricing for variable traffic patterns. This contrasts with committed rates, where customers pay a flat fee for guaranteed capacity up to the agreed limit, often at a lower per-unit cost but without flexibility for bursts; overages in committed plans may incur additional fees, while burstable plans allow temporary exceedances without penalty up to the port speed.^[51]^[52]^[53] Volume discounts are standard in transit agreements, with larger commitments yielding lower per-Mbps rates; for instance, committing to higher levels like 10 Gbps can reduce costs compared to smaller ports starting at 1 Gbps minimums. Contracts typically span 1 to 3 years to lock in pricing stability, and higher committed data rates (e.g., 10% or more of port capacity) often qualify for reduced burstable overage charges. These terms encourage long-term partnerships, particularly among Tier 2 and Tier 3 providers seeking cost efficiencies from upstream Tier 1 carriers.^[1]^[54]^[55] Market trends show sustained price declines driven by fiber optic overbuilds and increased competition, with average IP transit rates dropping from approximately $675 per Mbps in 2000 to under $1 per Mbps by 2023 in mature markets. This erosion reflects expanded submarine cable capacity and terrestrial fiber deployments, enabling providers to offer more bandwidth at lower marginal costs. The global internet transit market was valued at $495 billion in 2023 and is projected to reach $673 billion by 2030, growing at a compound annual growth rate (CAGR) of 4.5%, fueled by rising data demands despite pricing pressures.^[56]^[50]^[57]

Alternatives

Peering

Peering refers to the mutual exchange of traffic between autonomous systems (ASes) without monetary settlement, allowing networks to directly interconnect and provide access to each other's customers on a reciprocal basis.^[5] This arrangement is typically pursued by networks of comparable size or between content providers and those serving end-users (eyeball networks), ensuring balanced mutual benefit.^[58] Peering can occur through public or private interconnections, with the former facilitated at shared facilities and the latter via dedicated links. Public peering involves multiple networks connecting at a common point, enabling efficient multilateral exchanges, while private peering establishes direct, point-to-point connections between two ASes for higher traffic volumes.^[59] Most peering is settlement-free, where neither party pays for traffic carried, though paid peering exists when one network compensates the other for disproportionate value or capacity.^[60] Networks establish peering based on specific criteria to maintain fairness and efficiency, including balanced traffic ratios ideally near 1:1 to avoid one-sided backhaul costs.^[5] Peers must agree not to use each other for transit to third parties, announcing only their own customer prefixes, and often require minimum traffic volumes or backbone capacity to justify the interconnection.^[60] These relationships leverage BGP protocols for routing information exchange.^[59] Compared to paid transit, peering offers significant advantages, including substantial cost savings by eliminating upstream provider fees and sharing only circuit expenses.^[58] It also reduces latency through shorter, direct paths and fewer intermediaries, while decreasing dependency on third-party networks for global reachability.^[61] A prominent example is Google's extensive peering program, where it directly interconnects with ISPs to deliver content to end-users, bypassing transit providers for optimized performance and cost efficiency.^[59]

Internet Exchange Points (IXPs)

Internet Exchange Points (IXPs) serve as neutral, physical infrastructure facilities where multiple autonomous networks interconnect to exchange Internet traffic directly, bypassing traditional transit providers. These points provide a shared Layer 2 switching fabric that enables multilateral peering, allowing participants to connect a single port and exchange traffic with numerous other networks efficiently. To simplify peering arrangements, many IXPs deploy route servers that act as centralized BGP peers, redistributing routes among members and reducing the need for individual bilateral BGP sessions between every pair of networks. This setup facilitates scalable interconnection without the overhead of managing hundreds of direct sessions.^[62]^[63] Prominent examples of IXPs include the Amsterdam Internet Exchange (AMS-IX), which connects over 800 member networks in Amsterdam and operates additional exchanges globally; DE-CIX in Frankfurt, with more than 1,000 participants handling peak traffic exceeding 18 Tbit/s (all-time peak 18.22 Tbit/s as of 2025); and Equinix Internet Exchange facilities, which span more than 40 markets worldwide and interconnect 2,334 autonomous systems across 1,947 organizations.^[64]^[65]^[66] As of 2025, the number of IXPs worldwide has grown to over 1,000, spanning more than 140 countries, reflecting the increasing demand for localized traffic exchange infrastructure.^[67] IXPs offer key benefits such as enhanced scalability for peering, as networks can expand connections without proportional increases in infrastructure; reduced costs compared to establishing dedicated bilateral links or relying on paid transit; and efficient traffic aggregation, which keeps data exchanges local and minimizes latency. These advantages promote a more resilient and cost-effective Internet ecosystem by enabling direct, settlement-free traffic flows among diverse participants.^[68]^[69] Technically, IXPs are built around high-capacity Layer 2 Ethernet switches forming the core switching fabric, with Virtual Local Area Networks (VLANs) used to separate different participant groups or services for security and management. Support for IPv6 is commonly integrated into the fabric, allowing seamless dual-stack operation alongside IPv4, while optional multicast capabilities enable efficient distribution of group-based traffic when required by members. As of 2023, IXPs collectively handle an estimated 20–30% of global Internet traffic, significantly reducing reliance on transit by localizing exchanges and optimizing routing paths.^[70]^[71]^[69]

Challenges and Future Directions

Operational and Economic Challenges

Internet transit faces significant operational challenges due to the relentless growth of the BGP routing table, which exceeded 1 million IPv4 entries by late 2024 and continued to grow throughout 2025, with projections indicating further expansion.^[72]^[73] This expansion strains router hardware, particularly ternary content-addressable memory (TCAM) limits, leading to processing delays, dropped route updates, and unintended traffic blackholing where packets are silently discarded.^[72]^[74] To mitigate these issues, network operators employ traffic engineering techniques such as route filtering, prefix summarization, and selective acceptance of partial tables from transit providers, preventing table overflow and maintaining routing stability.^[75] Economic pressures in Internet transit have intensified with surging bandwidth demands from video streaming and cloud services, which now account for over 80% of global Internet traffic.^[76] This growth drives up infrastructure costs for transit providers, as they upgrade to higher-capacity links like 400 GigE ports, which cost 3.3 times more per Mbps than 100 GigE equivalents despite overall price erosion.^[50] Depeering disputes exemplify these tensions; in 2010, Comcast demanded fees from Level 3 Communications due to a 5:1 traffic imbalance caused by Netflix's streaming load, leading to a temporary agreement under protest that highlighted imbalances in content delivery networks.^[77] Security vulnerabilities remain a core operational concern, with BGP hijacks enabling traffic redirection and data interception. In April 2018, a small ISP (eNet) hijacked an Amazon Route 53 prefix, diverting DNS queries for cryptocurrency wallet MyEtherWallet and enabling over $150,000 in thefts while briefly disrupting services for clients like Instagram.^[78] Such incidents underscore the protocol's trust-based flaws, prompting adoption of Resource Public Key Infrastructure (RPKI) for route origin validation; by 2025, over 50% of IPv4 and IPv6 routes are secured via RPKI.^[79] Regulatory issues, particularly net neutrality debates, further complicate transit economics. The U.S. FCC's 2015 Open Internet Order reclassified broadband as a Title II service, prohibiting paid prioritization and subjecting interconnection disputes—including transit agreements—to case-by-case review under just-and-reasonable standards, which influenced peering negotiations without imposing direct rate controls.^[80] These rules, in effect from 2015 to 2017 before repeal, aimed to curb discriminatory pricing but added compliance costs estimated at $11 billion annually, potentially passed to consumers.^[80] Transit dependency introduces single points of failure, contributing to network outages; surveys indicate that third-party provider issues, often involving transit disruptions or BGP errors, account for about 39% of significant downtime events.^[81]

Emerging Trends (Post-2020)

The global Internet transit market, valued at USD 495 billion in 2023, is projected to reach USD 673 billion by 2030, growing at a compound annual growth rate of 4.5%, primarily driven by the expansion of hyperscale data centers and increasing demand for high-bandwidth connectivity despite the rising prevalence of peering arrangements.^[57] This growth reflects the continued importance of transit in supporting backbone infrastructure, even as alternative models erode its dominance. Post-2020, traditional Internet transit has experienced a notable decline in its share of overall traffic, dropping to approximately 40–50% by 2025 due to the proliferation of direct peering agreements and content delivery networks (CDNs) that enable content providers to deploy their own infrastructure closer to end users, reducing reliance on upstream transit providers.^[82] For instance, in European markets, inbound ISP traffic was split roughly 54% transit and 44% private peering in late 2024, signaling a trend toward balanced or further reduced transit proportions globally as CDNs like those operated by major hyperscalers cache content locally.^[83] This shift is amplified by Internet exchange points (IXPs), which facilitate efficient traffic exchange without transit intermediaries. Integration of cloud and edge computing has further transformed transit dynamics, with services such as AWS Direct Connect providing dedicated private connections that bypass public Internet transit routes, offering lower latency and enhanced security for data transfer between on-premises networks and AWS cloud resources.^[84] Similarly, the adoption of IPv6 in transit networks has accelerated, reaching over 45% of global traffic by late 2025, enabling more efficient addressing and routing for the expanding Internet ecosystem.^[85] Complementing this, software-defined networking (SDN) has gained traction among IP transit providers for enabling dynamic routing and automation, as seen in implementations by Tier 1 carriers like NTT Communications, which leverage SDN to optimize multi-country connectivity and respond to variable traffic demands.^[44] Artificial intelligence (AI) and machine learning (ML) are increasingly applied to optimize Internet transit for surges from 5G and Internet of Things (IoT) deployments, using predictive analytics for real-time traffic management and resource allocation to minimize congestion and improve efficiency.^[86] In 5G-enabled networks, AI-driven tools monitor traffic patterns and dynamically adjust routing, supporting the orchestration of IoT device connectivity while handling exponential data growth.^[87] These advancements collectively position Internet transit to evolve into a more agile, integrated component of the broader digital infrastructure.

References

[1]
Definition of Internet Transit - The Internet Peering Playbook
Internet Transit is the business relationship whereby an Internet Service Provider provides (usually sells) access to the global Internet.
[2]
What is IP Transit? - Colt Technology Services
Jul 11, 2025 · IP transit is a paid service where a network pays for access to the entire internet through another network's infrastructure, whereas peering is ...
[3]
What is an Internet IP Transit Provider? - ThousandEyes
Transit providers are ISPs that offer commercial services to carry traffic from ISP or customer networks and "transit" it to the rest of the Internet.
[4]
IP Transit and the Tiers of Transit Providers - Noction
Apr 12, 2022 · The role of a transit provider, also called an upstream provider, is to connect a customer's network or downstream ISP to the global Internet.
[5]
[PDF] Peering 101 - NANOG
•1) Internet Transit. •2) Internet Peering. •3) Internet Peering Ecosystem ... 3) Definition of an “Open” Peering Policy: 4 ...
[6]
Accepting and Announcing Routes - TechLibrary - Juniper Networks
A transit provider typically does not send a default route, although they can usually do so on request. In addition to sending you their full routing table, a ...
[7]
Global Tier 1 IP Networks: Everything You Need To Know - GTT
Jul 26, 2023 · They buy transit from Tier 1 ISPs and provide transit to Tier 3 ISPs. While Tier 1 ISPs cover enterprise-scale global clients for dedicated ...
[8]
Understanding Tier 1 ISPs: A Comprehensive Guide - 123NET
Mar 28, 2024 · Tier 1 Internet Service Providers (ISPs) represent the pinnacle of the internet's hierarchy, providing a backbone that ensures global connectivity.
[9]
Model for Inter-Domain Routing | CS 168 Textbook
At the very top of the hierarchy, there are Tier 1 autonomous systems, which have no providers (no incoming edges). Every Tier 1 AS has a peering relationship ...Missing: default- dependency
[10]
World - ASN statistics by number
Total number: 119 760 ; Oct 1, 2022, 103,013 ; Nov 1, 2022, 103,340 ; Dec 1, 2022, 103,692 ; Jan 1, 2023, 104,249.
[11]
Internet Statistics - bgp.he.net
Autonomous Systems with IPv4 Announcements Observed: 79,147. Autonomous Systems with IPv6 Announcements Observed: 37,688 IPv4 Prefixes Observed: 1,227,061<|control11|><|separator|>
[12]
The Role of Transit IP in Global Internet Infrastructure - LARUS
Apr 9, 2025 · It facilitates the exchange of data packets between different Internet Service Providers (ISPs) and networks, enabling global connectivity.
[13]
Understanding Internet Transit: A Comprehensive Guide - 123NET
Mar 12, 2024 · Internet transit functions through agreements between ISPs, where a provider agrees to carry traffic across their network. This service is ...
[14]
A Brief History of the Internet - Internet Society
Leonard Kleinrock at MIT published the first paper on packet switching theory in July 1961 and the first book on the subject in 1964. Kleinrock convinced ...
[15]
RFC 1093: NSFNET routing architecture
1. Routing Overview The new NSFNET backbone forms the core of the overall NSFNET, which connects to regional networks (or regional backbones) as well as to peer ...
[16]
Final report on TCP/IP migration in 1983 - Internet Society
Sep 15, 2016 · The immediate impact of TCP/IP adoption was a huge increase in the available address space, as 32 bits allows for approximately 4 billion hosts.
[17]
ARPANET Adopts TCP/IP - IEEE Communications Society
ARPANET architects decide to replace the existing Network Control Program (NCP) with TCP/IP on all ARPANET hosts. By June 1983, every host was running TCP/IP.
[18]
The History of TCP/IP
The migration of the ARPANet to TCP/IP was officially completed on January 1, 1983, when the new protocols were permanently activated. In March 1982, the US ...
[19]
NSFNET 20th Anniversary | North American Network Operators Group
When it was originally launched in 1985, the NSFNET connected six super computer centers across the United States via a 56Kbps backbone network. In late 1987, ...
[20]
[PDF] A Partnership for High-Speed Networking Final Report 1987-1995
Over time, the NSFNET program would include funds to construct, operate and maintain the NSFNET backbone network, to support the regional and midlevel networks, ...
[21]
[PDF] The NSFNET Backbone Network - NTP.org
The NSFNET Backbone Network interconnects six supercomputer sites, several regional net- works and ARPANET. It supports the DARPA Internet protocol suite and ...
[22]
NSF Shapes the Internet's Evolution - National Science Foundation
Jul 25, 2003 · To handle the increasing data traffic, the NSFNET backbone became the first national 45-megabits-per-second Internet network in 1991. The ...Missing: deployment | Show results with:deployment
[23]
Internet Ascendant, Part 2: Going Private and Going Public
Oct 22, 2020 · NSFNET would shut down in the spring of 1995, and its assets would revert to IBM and MCI. The regional networks could continue to operate, with ...Missing: transit | Show results with:transit
[24]
Internet Exchanges: Policy-Driven Evolution - CAIDA
The end of the transition from the federally supported Internet to a privatized Internet was marked by the decommissioning of the NSFNET backbone in 1995. Peer ...Missing: UUNET | Show results with:UUNET
[25]
[PDF] Retiring the NSFNET Backbone Service: Chronicling the End of an Era
We begin by taking a brief look at the history of what was the world's largest and fastest network for research and education. A Brief History of the NSFNET.
[26]
[PDF] Internet traffic growth: Sources and implications
Nov 30, 2000 · The most popular and extremely misleading myths of the dot-com and telecom bubbles was that “Internet traffic doubles every 100 days” (or 3 ...
[27]
Genuity Agrees to Sell Assets to Level 3 - Los Angeles Times
Nov 28, 2002 · With the Genuity deal, Level 3, a high-speed communications company, would gain Internet access services, more than 3,000 customers such as ...
[28]
Monopoly.com: Will the WorldCom-MCI Merger Tangle the Web?
The company will own four major Internet backbones: MCI, UUNet, ANS, and Compuserve (the latter three are already owned by WorldCom). It will administer five ...
[29]
https://datatracker.ietf.org/doc/html/rfc4271
[30]
RFC 4271 - A Border Gateway Protocol 4 (BGP-4) - IETF Datatracker
RFC 4271 defines BGP-4, an inter-Autonomous System routing protocol that exchanges network reachability information and supports CIDR.
[31]
RFC 1771: A Border Gateway Protocol 4 (BGP-4)
The Border Gateway Protocol (BGP) is an inter-Autonomous System routing protocol. It is built on experience gained with EGP as defined in RFC 904.
[32]
Active BGP entries (FIB) - BGP potaroo.net
... 2025 (UTC+1000). Active BGP entries (FIB). Table Size Metrics. The trend of the size of the BGP Forwarding Table (FIB). Also the underlying BGP Routing Table ( ...
[33]
BGP Routing: An In-Depth Tutorial and Examples - Kentik
Jul 3, 2025 · BGP is a policy-based routing protocol that ensures the efficient and reliable transfer of data packets across different autonomous systems (AS) ...
[34]
What you should know about BGP's LOCAL_PREF - Noction
Oct 7, 2015 · The LOCAL_PREF (local preference) is the first attribute a Cisco router looks at to determine which route towards a certain destination is the “best” one.
[35]
Understand BGP MED Attribute - Cisco
This document describes the Border Gateway Protocol (BGP) MED Attribute when it crosses over an AS boundary by implementation in different scenarios.
[36]
From Idle to Established: BGP states, BGP ports and TCP interactions
Apr 9, 2024 · Unlike interior gateway protocols (IGPs), BGP manually configures neighbors and uses TCP port 179 for establishing reliable connections.
[37]
BGP Essentials: Non-transit AS - ipSpace.net blog
Apr 8, 2008 · Here's the BGP configuration you should use on Cisco IOS: apply AS-path access-list to outbound updates with neighbor filter-list command.
[38]
BGP Filtering Best Practices - eBook - Noction
BGP filtering is used to control prefixes that are received and advertised to BGP peers. Filtering is critically important at Tier1, Tier2 and Tier3 levels.
[39]
BGP full-routes vs partial-routes vs default-route - AboutNetworks.net
Jul 18, 2018 · Uses-cases and examples of different BGP architectures: default-route versus full-routes versus partial-routes BGP peerings.
[40]
Difference between default route, partial and full routing table for a ...
Feb 13, 2018 · A partial routing table usually includes all the routes for customers of the ISP, and a default route for everything else.
[41]
https://www.consp.com/ip-internet-protocol-transit
[42]
Cross Connects vs. Virtual Connects - The Equinix Blog
Jan 19, 2023 · Cross connects are point-to-point cable links, while virtual connections are software-defined, allowing multiple connections on a single port.
[43]
Not all IP transit providers are created equal! How to choose the ...
Providers should offer service level agreements (SLAs) that guarantee a minimum level of uptime, commonly 99.99% or higher. Enhanced Security: Quality ...
[44]
IP (Internet Protocol) Transit - Chase Freedom of Traffic
IP Transit offers guaranteed performance levels in critical areas such as throughput, packet loss and latency with 100% availability.Missing: effort | Show results with:effort
[45]
Configure the BGP Maximum-Prefix Feature - Cisco
By default, this feature allows a router to bring down a peer when the number of received prefixes from that peer exceeds the configured Maximum-Prefix limit.
[46]
Understanding BGP Communities | Noction
BGP communities are labels attached to BGP routes, some with pre-defined meanings, and user-defined for custom routing policies.
[47]
Tier 1 ISPs: A Comprehensive Guide to Global Internet Connectivity
Apr 24, 2025 · Significant Cost Savings: By avoiding transit fees, Tier 1 ISPs like Lumen Technologies save enterprises millions annually. For example, a ...
[48]
Internet Service Provider 3-Tier Model - ThousandEyes
A Tier 2 ISP is a service provider that utilizes a combination of paid transit via Tier 1 ISPs and peering with other Tier 2 ISPs to deliver Internet traffic to ...
[49]
Tier 1 vs Tier 2 vs Tier 3 ISPs Explained: The Complete Guide for IT ...
Aug 26, 2025 · Definition: A Tier 1 ISP can reach every other network on the Internet via settlement-free peering—never paying anyone for transit. Technical ...Missing: non- | Show results with:non-
[50]
[PDF] evolution-of-internet-interconnection.pdf - Charles River Associates
The Internet has evolved from a ''hierarchy''—in which interconnection was achieved by having Internet Service Providers (ISPs) purchase transit services ...Missing: IXPs | Show results with:IXPs
[51]
The European Network Usage Fees proposal is about much more ...
May 8, 2023 · ... [prices on offer for] 100 GigE [IP transit services in Europe] were $0.06 per Mbps per month.” These prices are consistent with what ...<|separator|>
[52]
Latin American Pricing Takeaways From ITW - TeleGeography Blog
Jun 7, 2023 · According to our Wavelengths Network Pricing Database, in Q1 2023, the weighted median 100 Gbps price on Miami-Sao Paulo ($18,000) was just ...
[53]
95th percentile and other bandwidth metering methods | Noction
The 95th percentile is a method of metering bandwidth usage that allows clients to slightly burst over their committed rate.
[54]
95th Percentile Bandwidth Metering Explained | Auvik
Sep 20, 2024 · 95th percentile bandwidth metering is a method to determine usage 95% of the time. Learn how to calculate & do capacity planning.Missing: transit | Show results with:transit
[55]
Inter.link A Buyer's Guide to IP Transit Pricing
Mar 6, 2024 · Committed versus burst bandwidth: The more Committed Data Rate (CDR) you are willing to commit to, the lower your base rate and the lower the ...
[56]
IP Transit: Services, Providers & Pricing (Beginner's Guide)
Oct 6, 2025 · Pricing models and contracts: ... Contracts for these services are typically time-bound (12-36 months) and may offer volume discounts depending on ...Missing: minimums | Show results with:minimums
[57]
Negotiating IP Transit Pricing: Tips For Getting The Best Deal
Negotiate contracts with your IP Transit service provider. Cost savings can be accessed through discounts on volume, long-term contracts and bundle services.
[58]
Internet Transit Prices - Historical and Projections
The unmistakable Transit Pricing trend is down, with an average decline of 61% from 1998 to 2010 as shown in the graph below. Internet Transit Prices ( ...
[59]
IP Transit Pricing in 2025: More Competition, More Price Erosion
Sep 8, 2025 · In Q2 2025, the lowest 100 GigE prices on offer in the most competitive markets remained steady at $0.05 per Mbps per month. The lowest for 10 ...Missing: 2023 | Show results with:2023
[60]
Global Internet Transit Market: Industry Analysis and Forecast
Internet Transit market was valued at USD 495 Bn in 2023 is expected to reach USD 673 Bn by the end of 2030 at a CAGR of 4.50% from 2024 to 2030.
[61]
[PDF] Internet peering and settlements | APNIC
These two forms of interconnection, namely the customer/provider relationship and the SKA peer relationship, form the basis of the entire set of connections ...Missing: definition types criteria advantages
[62]
[PDF] Internet Transit, Peering & the Variants | APNIC Academy
Aug 26, 2021 · v1.3. Peering in General. • ASes are interconnected/peered at Internet exchanges points (IXPs) or privately. • Interconnection/peering is among ...Missing: definition criteria
[63]
None
### Summary of Internet Peering from NANOG51 Presentation
[64]
[PDF] A Guide to Peering on the Internet - LaFibre.info
Jan 30, 2011 · • Peer – Two networks who get together and agree to exchange. traffic between each others' networks, typically for free. 3.Missing: criteria | Show results with:criteria
[65]
Internet Exchange Point Overview - Documentation - Juniper Networks
An Internet Exchange Point (IXP) is a Layer 2 network that facilitates interconnection between ISPs using BGP to exchange routing information.
[66]
RFC 7948: Internet Exchange BGP Route Server Operations
RFC 7948 describes how BGP route servers reduce overhead at IXPs by redistributing BGP routes, enabling multilateral interconnection and reducing the number of ...
[67]
IXP Details: Amsterdam Internet Exchange - AMS-IX
IXP Details: Amsterdam Internet Exchange - AMS-IX · Country: Netherlands · City: Amsterdam · Members: 823 (View members) ...
[68]
DE-CIX Frankfurt - BGPView
Members: 1014. Tech Email: support@de-cix.net. Tech Phone: +4969173090211. Policy Phone: sales@de-cix.net. Policy Phone: +4969173090212. DE-CIX Frankfurt ...Missing: 2023 | Show results with:2023
[69]
Equinix Internet Exchange
### Summary of Equinix Internet Exchange Participants (circa 2023)
[70]
[PDF] Introduction to Internet Exchange Points
• 700+ IXPs worldwide across 130+ countries. • IXPs serve as the backbone of ... lookup time for its members of the IXP in 2023 through the +Raices ...
[71]
Importance of Internet Exchange Points | LSIX News
Jan 22, 2020 · Scalability: An IXP can provide multiple ports of varying capacities. It enables the user to scale capacity if and when needed. Cost-Efficient:
[72]
[PDF] Promoting the Use of Internet Exchange Points: A Guide to Policy ...
The primary role of an IXP is to keep local traffic local and reduce the costs associated with traffic exchange be- tween Internet providers. The case for IXPs ...
[73]
RFC 5963 - IPv6 Deployment in Internet Exchange Points (IXPs)
Oct 14, 2015 · This document provides guidance on IPv6 deployment in Internet Exchange Points (IXPs). It includes information regarding the switch fabric configuration, the ...
[74]
Netnod's IXP Architecture
This page explains the technical setup at Netnod IXes and provides important connection details for customers. Technical overview. Netnod's Internet exchange ...
[75]
BGP in 2024 - APNIC Blog
Jan 6, 2025 · In terms of advertised prefixes, the size of the routing table grew by 53,000 entries or 6%. The number of root prefixes increased by 13,000 ...
[76]
The One Million Route Problem | What It Is & How to Survive It
May 22, 2025 · Your BGP sessions may remain up, but routes will silently be dropped. Traffic blackholes, routing loops, and widespread instability can occur.
[77]
Blackhole route for BGP Summarization - the Fortinet Community!
Dec 19, 2024 · By default, BGP only advertises prefixes that are present in the routing table. This ensures that only valid and reachable routes are propagated ...
[78]
The Future of IP Transit: Trends in Network Capacity and Global ...
Feb 7, 2025 · In this guide, we'll explore the key trends shaping the future of IP transit, including increasing bandwidth needs, next-generation routing ...
[79]
Level 3 vs. Comcast: More Than A Peering Spat?
Nov 29, 2010 · A dispute between Level 3 and Comcast appears to have been precipitated by Level 3's recent addition of online video service Netflix as a major ...
[80]
Anatomy of a BGP Hijack on Amazon's Route 53 DNS Service
Sitting in the XLHost data center was a fake DNS server that selectively answered queries for MyEtherWallet.com. All other requests were silently discarded.
[81]
Adoption of RPKI/ROV Security Protocol Progressing Very Quickly
Mar 18, 2025 · More than half of both the IPv4 routes and the IPv6 routes in the BGP internet routing system are now secured with RPKI.
[82]
[PDF] Federal Communications Commission FCC 15-24 Before the ...
Mar 12, 2015 · ... Internet. 2. Four years ago, the Commission adopted open Internet rules to protect and promote the. “virtuous cycle” that drives innovation ...<|separator|>
[83]
Significant Telecom Network Outages 2023-2024 - OPT/NET BLOG
Apr 16, 2025 · The 2023 Uptime Resiliency Survey highlights that configuration failures (45%) and third-party provider issues (39%) are the leading causes.
[84]
Content providers and the deployment of Internet infrastructure
This paper documents the growing role that content providers play upstream in the global internet supply chain.
[85]
[PDF] Arcep Barometer of data interconnexion in France (July 4, 2025)
Jul 4, 2025 · In the second half of 2024, inbound traffic to ISPs' networks was split chiefly between transit (around 54.2%) and private peering (around 44.4 ...
[86]
AWS Direct Connect - AWS Documentation
AWS Direct Connect links your internal network to an AWS location via fiber-optic cable, creating virtual interfaces to AWS services, bypassing internet ...Direct Connect components · Network requirements · Supported Direct Connect...
[87]
IPv6 Adoption - Google
The graph shows the percentage of users that access Google over IPv6. Native: 45.26% 6to4/Teredo: 0.00% Total IPv6: 45.26% | Oct 30, 2025.
[88]
Revolutionizing Telecom with AI: Key Solutions for 5G and IoT - Subex
Sep 18, 2024 · Discover how AI is revolutionizing the telecom industry by optimizing 5G and IoT networks, enhancing customer experiences, and preventing ...
[89]
Leveraging Machine Learning and Artificial Intelligence for 5G
Embedding ML algorithms and AI into 5G networks can enhance automation and adaptability, enabling efficient orchestration and dynamic provisioning of the ...