ISPs and Tiers: The Business of Internet Connectivity

When you pay your monthly Internet bill, you're participating in a vast economic ecosystem invisible to most users. Your payment to a local ISP is just the visible surface of a multi-layered commercial structure where networks pay for access to other networks, negotiate traffic exchange agreements, and compete for the most efficient paths to global destinations.

Understanding this ISP hierarchy isn't just academic curiosity—it explains why your connection to some sites is fast and others slow, why certain outages cascade across the Internet, and why network engineering decisions have profound business implications. For network engineers, understanding ISP economics is essential for negotiating connectivity, troubleshooting inter-domain issues, and optimizing application performance.

The Internet's commercial structure evolved organically from its academic origins. When NSF decommissioned NSFNET in 1995 and the backbone became commercial, no central authority planned the resulting structure. Instead, market forces created a hierarchical ecosystem of Internet Service Providers.

What Is an ISP?

An Internet Service Provider (ISP) is an organization that provides Internet connectivity services. This broad definition encompasses vastly different entities:

• National backbone operators with global infrastructure
• Regional carriers serving specific geographic areas
• Local access providers connecting homes and businesses
• Hosting companies providing server connectivity
• Content providers with their own network infrastructure

Why Hierarchy Emerged:

No single network can connect directly to every other network—there are over 100,000 autonomous systems globally. The costs of maintaining that many direct connections would be prohibitive. Instead, networks form hierarchical relationships:

1. Small networks pay larger networks for access to the global Internet
2. Larger networks aggregate traffic from many smaller networks
3. The largest networks exchange traffic directly with each other

This creates an implicit hierarchy, though the boundaries are fuzzy and the relationships complex. The commonly used tier classification attempts to categorize networks by their position in this hierarchy.

Tier 1 networks sit at the apex of the Internet hierarchy. They possess a defining characteristic: they can reach every destination on the Internet without paying any other network for transit.

The Defining Criterion:

A network qualifies as Tier 1 if it:

1. Has settlement-free peering agreements with every other Tier 1 network
2. Can reach 100% of Internet destinations through its own network and peering relationships
3. Never purchases transit from any other network

This is a stringent requirement. Most networks, regardless of size, purchase transit from at least one upstream provider. Only a small number of networks have the global reach, traffic volumes, and business relationships to achieve true Tier 1 status.

Why Tier 1 Status Matters:

Tier 1 networks hold structural power in the Internet ecosystem:

• No upstream dependencies — They cannot be disconnected by a transit provider's business decision
• Peering leverage — Other networks need them more than they need any individual peer
• Traffic aggregation — They carry traffic from thousands of downstream networks
• Infrastructure scale — Their backbone investments enable global reach

The Peering Mesh:

Tier 1 networks peer with each other at settlement-free terms—neither party pays the other. This works because traffic is roughly balanced: Network A's customers want to reach Network B's customers, and vice versa. The mutual benefit justifies the interconnection costs.

These peering agreements form a full mesh: every Tier 1 can reach every other Tier 1 directly. This mesh provides the Internet's ultimate backstop—if a network can reach any Tier 1, it can reach every other network.

Tier 2 networks constitute the vast middle tier of ISPs. They cannot reach the entire Internet through peering alone and must purchase transit from Tier 1 providers for at least some destinations.

Characteristics of Tier 2 Networks:

• Mix of peering and transit — They peer where possible to reduce costs and improve performance, but pay transit for unreachable destinations
• Regional or specialized focus — Often strong in specific geographies or market segments
• May sell transit — Often provide upstream connectivity to smaller Tier 3 networks
• Significant infrastructure — Operate substantial network infrastructure, just not globally comprehensive

Tier 2 Examples:

Tier 2 networks include:

• National carriers — Large ISPs with extensive domestic infrastructure (British Telecom, Deutsche Telekom, NTT East/West)
• Regional specialists — Networks strong in specific regions but lacking global reach
• Large cable companies — Cable MSOs with significant backbone infrastructure (formerly Tier 3 companies like Cox, Charter)
• Telecommunications providers — Traditional telcos that have built IP networks

The Transit Decision:

Tier 2 networks constantly optimize their mix of peering and transit:

More peering means:

• Lower transit costs
• Better control over routing
• Improved latency (direct paths)
• Increased operational complexity

More transit means:

• Simpler operations
• More predictable costs (though higher)
• Dependence on transit provider performance
• Less routing flexibility

Tier 3 networks are purely customer-focused ISPs that buy all their upstream connectivity from larger providers. They don't sell transit to other networks—their business is providing Internet access to end users and businesses.

Tier 3 Characteristics:

• Pure access providers — Focus on the "last mile" connecting homes and businesses
• Transit-only upstream — Purchase all upstream bandwidth from Tier 1 or Tier 2 providers
• Local or regional scope — Serve specific cities, regions, or market niches
• High customer count, modest traffic — Many individual connections, each with relatively small bandwidth needs

Types of Tier 3/Access ISPs:

Multi-Homing for Reliability:

Larger Tier 3 networks often multi-home—purchasing transit from multiple upstream providers:

Benefits:

• Redundancy — If one upstream fails, traffic routes through others
• Performance — Can route different destinations via the best-performing upstream
• Negotiating leverage — Competition between upstreams produces better pricing
• Load balancing — Distribute traffic across providers to optimize costs

Complexity:

• BGP configuration — Must run BGP and manage multiple upstream sessions
• Address space — Typically need provider-independent IP addresses
• Routing policy — Must decide how to route outbound traffic and advertise inbound
• Operational overhead — More relationships mean more coordination during issues

Enterprise Networks:

Large enterprises often function like Tier 3 ISPs internally:

• Purchase transit from multiple upstream providers
• Operate private backbone networks connecting offices globally
• May peer directly with frequently-accessed content providers
• Don't sell transit but have substantial infrastructure

Enterprise network engineers face similar decisions about multi-homing, peering, and transit optimization as commercial access ISPs.

Transit is the fundamental commercial relationship in Internet interconnection: one network pays another for access to the global Internet.

How Transit Works:

When Network A purchases transit from Network B:

1. Network B advertises all routes — B provides A with routes to every destination B can reach (essentially, the entire Internet)
2. Network B carries A's traffic — B forwards A's packets toward their destinations, whether those destinations are B's customers, B's peers, or reached via B's own transit
3. Network B advertises A's routes — B announces A's IP prefixes to B's peers and upstreams, ensuring return traffic reaches A
4. Network A pays Network B — Typically based on bandwidth committed or used

Transit Pricing Models:

95th Percentile Billing:

The most common transit pricing model is 95th percentile billing:

1. Traffic is sampled every 5 minutes (288 samples/day)
2. At month end, samples are sorted from highest to lowest
3. The top 5% of samples are discarded
4. The highest remaining sample = billable bandwidth

This model allows occasional traffic bursts without penalty while pricing based on sustained usage. A network that normally uses 1 Gbps but spikes to 5 Gbps briefly still pays based on approximately 1 Gbps.

Transit Prices:

Transit prices have declined dramatically over decades:

• 1998: ~$1,200 per Mbps/month
• 2008: ~$20 per Mbps/month
• 2018: ~$0.50 per Mbps/month
• 2024: As low as $0.10-0.30 per Mbps/month for large volumes

These price declines reflect technological improvements, competition, and optimization. However, prices vary significantly by geography—transit in developing regions can cost 10-100x more than in competitive markets like Amsterdam or Ashburn.

Peering is a non-transitive relationship where two networks exchange traffic destined for each other's customers—without payment changing hands (in settlement-free peering) or with payment reflecting traffic imbalance (in paid peering).

Peering vs Transit:

The fundamental difference:

Transit: Network A receives routes to the entire Internet and pays Network B for this access.

Peering: Networks A and B exchange routes only to their own customers. Neither gains access to the other's upstream or peers.

Types of Peering:

Settlement-Free Peering: Neither party pays the other. This is the ideal for both parties when traffic is balanced and interconnection is mutually beneficial.

Paid Peering: The network sending more traffic (typically content-heavy) pays the network receiving more traffic (typically access-heavy). This addresses traffic imbalance while maintaining direct interconnection.

Public Peering: Occurs at Internet Exchange Points where many networks share common switching infrastructure. Lower cost per peer but less control and capacity.

Private Peering: Direct fiber connections between two networks' routers. Higher cost but dedicated capacity and complete control.

Peering Policies:

Networks establish peering policies defining their requirements for peering relationships. Common criteria include:

• Minimum traffic volumes — Must exchange at least X Gbps to justify operational costs
• Geographic presence — Must peer at multiple locations for redundancy
• Traffic ratios — Traffic must be reasonably balanced (e.g., no more than 2:1 imbalance)
• Network type — May only peer with similar network types (not competitors, content-only, etc.)
• 24/7 NOC — Must have operational capability to resolve issues
• Route stability — Must maintain stable routing without excessive churn

Policies range from open (peer with anyone meeting minimal criteria) to selective (peer only with networks offering significant value) to restrictive (peer with almost no one).

The traditional ISP hierarchy is evolving rapidly. Several trends are reshaping how the Internet's business relationships work:

The Rise of Content Networks:

Hyperscale content providers (Google, Meta, Netflix, Amazon, Microsoft) have built massive global networks that rival or exceed traditional Tier 1 carriers:

• Google: Global fiber backbone, substantial submarine cable ownership, edge caching everywhere
• Meta: Significant submarine cable investments, global edge network
• Netflix: Open Connect CDN deployed inside thousands of ISP networks
• Amazon/Microsoft: Global cloud infrastructure with extensive private backbone

These content networks don't sell transit—they're purpose-built to deliver their own traffic. But their scale makes them powerful actors in interconnection negotiations.

Flattening of the Hierarchy:

Traditional tier classification assumed strict hierarchy, but modern reality is flatter:

• CDNs peer everywhere — Content delivery networks bypass transit entirely through pervasive peering
• IXPs multiply — More exchange points enable more direct peering relationships
• Remote peering — Technology allows peering across distance without local presence
• Content sources peer directly with access networks — Cutting out intermediate transit

This flattening reduces the power of traditional transit providers while increasing the importance of peering and direct interconnection.

Consolidation Trends:

The ISP industry has seen significant consolidation:

• Backbone consolidation — Major acquisitions (Level 3 + CenturyLink, GTT + multiple acquisitions)
• Access provider consolidation — Cable and telco mergers creating regional monopolies
• Vertical integration — Access providers acquiring content (AT&T + WarnerMedia, later divested)

Consolidation concentrates power but can also create interconnection complications when a merger suddenly changes competitive dynamics.