Fiber Providers & Carriers

Introduction Back in 2009, I watched a major financial services company lose $2.3 million in a single afternoon because both their "redundant" fiber paths came into the building through the same conduit.

A contractor's backhoe taught them an expensive lesson about carrier diversity.

That incident changed how I approach fiber connectivity discussions with clients-and it should inform how you think about your data center's lifeline to the outside world.

Fiber providers and carriers represent your data center's umbilical cord to the internet.

Without them, you're operating a very expensive space heater.

The relationship between dark fiber and lit fiber, between carriers and last-mile providers, shapes everything from your network architecture to your disaster recovery planning.

The difference between Tier I carriers like Lumen and regional fiber providers isn't just marketing-it affects your latency, your redundancy options, and your negotiating position when renewals come around.

This lesson will equip you with the knowledge to evaluate fiber providers, understand when dark fiber makes financial sense versus lit services, and architect carrier diversity that actually protects your operations.

These aren't theoretical concepts-they're decisions that determine whether your data center survives real-world failures.

The Major Players in Fiber Infrastructure The fiber provider landscape has consolidated dramatically over the past decade, but understanding who owns what infrastructure matters when you're planning carrier diversity.

Zayo Group controls approximately 130,000 route miles of fiber in North America and Europe, making them the largest dark fiber provider I regularly work with.

When AWS built their US-EAST-1 region in Northern Virginia, Zayo's existing fiber infrastructure was a major consideration-hyperscalers don't build where fiber doesn't exist.

Lumen Technologies (formerly CenturyLink after acquiring Level 3 Communications) operates roughly 450,000 route miles globally.

They're one of the true Tier I carriers, meaning they don't pay for transit to reach any part of the internet.

Their submarine cable investments and intercontinental routes make them particularly valuable for facilities serving global applications.

I've seen Lumen's network redundancy save a streaming provider during the 2021 fiber cuts in the Midwest-traffic rerouted automatically because of their mesh topology.

Crown Castle Infrastructure specializes in metropolitan fiber networks and connects approximately 70,000 cell towers and small cells across major US markets.

Their strength lies in dense metro builds, making them critical for edge computing deployments.

Digital Realty's Portland facility leverages Crown Castle's metro rings to provide diverse connectivity options to multiple carrier hotels downtown.

Regional specialists also play crucial roles.

Cogent Communications focuses on providing cost-effective bandwidth in major metropolitan areas, though their network architecture prioritizes economics over redundancy.

Windstream operates extensive fiber in secondary markets across the southern US.

CoreSite's network campuses typically connect to 10-15 different fiber providers, creating genuine diversity options.

Provider	Route Miles	Primary Strength	Typical Use Case
Zayo Group	~130,000	Dark fiber availability	Inter-datacenter private networks
Lumen	~450,000	Global reach, Tier I status	International connectivity
Crown Castle	Metro-focused	Dense metropolitan builds	Edge compute, low-latency
Cogent	Metro-focused	Cost-effective bandwidth	Budget-conscious builds
Windstream	Regional US	Secondary market coverage	Regional operations

Dark Fiber vs.

Lit Fiber Economics Dark fiber means leasing the physical fiber strands themselves-you bring your own optics, routers, and management.

Lit fiber (or lit services) means the provider handles all the active equipment and delivers a working circuit at whatever speed you're paying for.

The decision between them fundamentally changes your cost structure and operational model.

I typically see the break-even point for dark fiber around 10Gbps sustained utilization over a 5-year term.

Running 10G over dark fiber might cost $2,500-3,500 monthly for the fiber lease, plus $8,000 in upfront optics and equipment.

Compare that to lit 10G service at $4,500-6,500 monthly with no equipment costs.

Simple math shows dark fiber wins after year two-but only if you have the operational capability to manage it.

Dark fiber advantages extend beyond pure economics.

When Microsoft builds inter-datacenter connections between their Azure regions, they use dark fiber almost exclusively because they need 400G and 800G wavelengths that many carriers don't even offer as lit services yet.

They control the upgrade cycle, the encryption, and the monitoring.

Equinix's dedicated fiber plant between their IBX facilities in major metros uses dense wavelength division multiplexing (DWDM) equipment over dark fiber strands, allowing them to scale capacity independently of carrier roadmaps.

The operational reality matters though.

Dark fiber requires staff who understand DWDM, can respond to fiber cuts, and manage optical power budgets.

Lit services shift that burden to the provider-they handle the monitoring, the truck rolls, and the equipment refreshes.

Smaller operators often choose lit services specifically because they can't justify a dedicated optical networking team.

Weather and physical events illustrate another critical difference.

When Hurricane Ida damaged fiber infrastructure in Louisiana in 2021, lit service customers waited for provider restoration crews.

Dark fiber customers with proper documentation could contract their own repair vendors and expedite restoration.

QTS's Atlanta-Metro facility experienced this during local construction damage-their dark fiber path was restored 18 hours faster because they controlled the repair authorization.

Understanding Carrier Diversity Requirements Carrier diversity sounds straightforward until you're standing in a meet-me room realizing three "different" carriers are all subleasing from the same underlying fiber owner.

True diversity requires understanding the physical path, the conduit routes, and the ownership chain of every fiber strand.

Physical path diversity means your fiber routes enter the building from geographically separated directions, ideally on different streets.

Equinix's SV5 facility in San Jose has fiber entrances on three different sides of the building, connected to diverse underground conduit systems.

When earthquake considerations enter the picture-which they do in California-you also need diverse routes that don't cross the same fault lines.

Carrier diversity extends beyond physical paths.

Having Zayo and Lumen circuits provides true provider diversity-different networks, different operational teams, different failure domains.

But if both carriers deliver service over fiber they're leasing from the local incumbent telco, you've just created a hidden single point of failure.

I've seen this exact scenario bring down supposedly redundant connections when the underlying fiber owner performed maintenance without notifying the retail providers.

Entry point diversity matters enormously for disaster scenarios.

Data center designs should include at minimum two physically separated fiber entry vaults, each connecting to different conduit systems.

Switch's TAHOE RENO 1 facility includes four separate fiber entry points, each connecting to different long-haul routes.

That architecture survived a 2018 incident where construction crews damaged one entry vault-three others maintained full connectivity.

Building truly diverse paths requires documentation and verification.

When CoreSite delivers carrier diversity to customers, they provide physical route maps showing actual conduit paths.

Smart operators request letter-of-agency access to carrier route databases to verify physical separation.

Digital Realty's 365 Main facility in San Francisco maintains detailed path diversity documentation because customers in financial services must prove regulatory compliance.

The cost implications are real.

Single-path connectivity might run $3,000 monthly.

Adding a second diverse path typically costs 80-100% of the primary path cost-not 50% like some vendors claim.

Facebook's data center network maintains at minimum three diverse long-haul paths from each facility, recognizing that carrier diversity represents insurance, not duplication.

Last-Mile Challenges and Metro Fiber Options The "last mile" problem-getting fiber from carrier points of presence (PoPs) to your actual data center-creates surprising bottlenecks.

Metro fiber availability determines which facilities can support high-bandwidth requirements.

Crown Castle and Zayo dominate metro fiber in major markets, but coverage gaps exist even in cities like Dallas and Atlanta.

Building-specific fiber availability varies wildly.

Equinix's carrier-dense facilities typically connect to 50+ carriers on-net.

Smaller colocation providers might offer 5-10 carriers, requiring costly last-mile extensions for additional diversity.

I've quoted last-mile fiber extensions ranging from $35,000 for a half-mile build to $280,000 for a complex urban route requiring boring under highways.

Lead times for new fiber extensions catch many operators off-guard.

Standard last-mile builds take 60-120 days in established markets.

Challenging routes involving railroad crossings, environmental permits, or multiple municipalities stretch to 6-9 months.

Google's expansion into its Council Bluffs, Iowa facility required 8 months of fiber construction because existing metro infrastructure didn't support their bandwidth requirements.

Metro fiber rings provide elegant solutions in mature markets.

Zayo operates metro rings in 50+ US cities, allowing customers to connect at any point on the ring.

This architecture provides path diversity and flexibility-if one segment fails, traffic routes around the ring.

Digital Realty leverages these rings to offer diverse connectivity between their facilities in markets like Chicago and Los Angeles.

Submarine cable landing stations represent unique last-mile scenarios.

Equinix's facilities in Ashburn, Virginia sit at the terminus of multiple transatlantic submarine cables.

The last few kilometers from ocean to data center required specialized construction techniques and involved multiple regulatory agencies.

Facilities near these landing points command premium pricing because the last-mile infrastructure already exists.

Service Level Agreements and Performance Commitments SLAs define what happens when fiber connectivity fails-and fiber does fail.

Realistic SLAs separate professional providers from optimistic amateurs.

Standard carrier SLAs promise 99.9% uptime, which mathematically allows 8.76 hours of downtime annually.

Premium SLAs reach 99.99% (52.56 minutes yearly) or even 99.999% (5.26 minutes yearly), but cost increases aren't linear-each additional nine roughly doubles the price.

Mean Time To Repair (MTTR) commitments matter more than uptime percentages for many applications.

Lumen's standard SLA promises 4-hour MTTR for fiber breaks.

Premium SLAs reduce this to 2 hours but require pre-positioned spare fibers and prioritized dispatch.

When Zayo's fiber was cut near Equinix's CH2 facility in Chicago in 2020, their 2-hour SLA commitment meant repair crews arrived within 40 minutes because they maintain forward-deployed repair kits in strategic locations.

Financial remedies for SLA violations rarely compensate for actual business impact.

Typical credits equal one month of service fees-maybe $5,000 for a 10G circuit.

If that outage cost your business $50,000 in lost revenue, the SLA credit feels insulting.

This reality drives carrier diversity requirements-SLA credits don't prevent outages, redundant paths do.

Packet loss and latency guarantees vary significantly by provider.

Cogent's standard service typically includes no packet loss guarantees.

Lumen and Zayo offer premium services with <0.1% packet loss commitments and specific latency targets.

AWS Direct Connect maintains sub-5ms latency between their facilities and local PoPs in most regions, backed by automatic traffic engineering that reroutes around congestion.

Practical Examples Example 1: Multi-Region Content Distribution Architecture A video streaming company I worked with needed to distribute content from their origin servers in CyrusOne's Phoenix facility to edge PoPs in 12 US markets.

They initially proposed using lit 100G circuits from a single carrier, which would have cost approximately $180,000 monthly with 60-day lead times per circuit.

We restructured the design using dark fiber between Phoenix and Los Angeles (their densest market), then lit services from the LA PoP to smaller markets.

The Phoenix-LA dark fiber cost $12,000 monthly for 10 strands, supporting 2 terabits of aggregate capacity using their owned DWDM equipment.

Lit 100G services from LA to secondary markets cost $4,500-8,500 monthly depending on distance.

Total monthly cost dropped to $98,000 while providing more capacity on the high-volume trunk route.

Physical diversity requirements added complexity.

We verified that their "diverse" paths between Phoenix and LA didn't share conduit at any point, requiring carrier route database access and physical site surveys at regeneration points.

The diverse path cost an additional 85% over the primary path-$22,000 monthly for both paths-but provided genuine protection.

When Union Pacific railroad maintenance damaged fiber near Yuma, Arizona in 2022, traffic failed over to the diverse path with zero customer impact. Example 2: Financial Services Carrier Selection Matrix A financial services firm selecting fiber providers for their new Atlanta facility needed to meet regulatory requirements for network diversity while minimizing latency to NYSE Mahwah.

They shortlisted five carriers, but actual evaluation revealed overlapping infrastructure that reduced true diversity.

We built a decision matrix evaluating: physical path separation (verified through site surveys), latency to key destinations (measured through test circuits), ownership chain (researched through corporate filings), and regional support capabilities (assessed through RFP responses and reference calls).

Zayo and Lumen emerged as truly diverse, with Crown Castle providing a third path for critical systems.

The final architecture used Zayo for primary connectivity (22ms to NYSE Mahwah), Lumen for diverse backup (23ms via different routing), and Crown Castle metro fiber for local connectivity to partner banks.

Total monthly cost reached $47,000 for three carriers versus $28,000 for a dual-carrier approach, but the regulatory compliance value and actual diversity justified the premium.

They documented physical separation exceeding 500 feet at all points, satisfying auditor requirements.

Common Misconceptions "More carriers equals better diversity" sounds logical until you examine actual infrastructure.

I've reviewed facilities claiming "20 carrier options" where 15 carriers were subleasing last-mile fiber from the same underlying provider.

Adding the sixth, seventh, and eighth carrier rarely improves diversity if they all use the same physical paths.

Focus on physical path diversity first, then add carrier diversity within those distinct paths.

Two truly diverse carriers beats five carriers sharing infrastructure. "Dark fiber is always cheaper at scale" oversimplifies the economics.

Dark fiber wins when you need sustained high bandwidth over stable routes with long contract terms.

But operational costs matter-you need optical engineering expertise, spare optics inventory, test equipment, and 24/7 monitoring capabilities.

Smaller operators choosing dark fiber to save money often discover operational costs exceed their savings.

Companies spending less than $50,000 monthly on connectivity rarely have enough scale to justify dark fiber's operational overhead.

Know your break-even point based on real operational costs, not just circuit pricing.

Summary & Key Takeaways

• Provider selection requires understanding physical infrastructure ownership, not just retail carrier names.

Verify that diverse carriers actually use separate physical paths through route database access and site surveys.

• Dark fiber becomes cost-effective around 10Gbps sustained utilization over 5+ year terms, but only when you have the operational capability to manage optical networks.

Lit services shift operational burden to the provider at the cost of flexibility and long-term economics.

• True carrier diversity demands physical path separation, diverse entry points, and independent ownership chains.

Single conduit routes create hidden single points of failure regardless of how many retail carriers you contract with.

• Major providers like Zayo, Lumen, and Crown Castle serve different needs-dark fiber availability, global reach, and metro density respectively.

Match provider strengths to your specific requirements rather than assuming bigger is always better.

• SLA financial remedies rarely compensate for actual outage costs-design for redundancy rather than relying on service credits.

The best SLA is the one you never need to invoke.

• Last-mile fiber availability determines site viability for high-bandwidth operations.

Budget 60-120 days and $35,000-280,000 for new fiber extensions in your site selection process.

Next Steps Study network topology design patterns to understand how carrier diversity integrates with your overall network architecture.

Examine meet-me room infrastructure and cross-connect strategies to maximize the value of multi-carrier connectivity.

Consider exploring submarine cable systems and intercontinental routing for global operations-the principles of carrier diversity scale from metropolitan to international deployments.