Library
Everything we have written, on one page
24 guides and research notes, grouped by the decision each one helps you make. They are free to read and they do not require an account. Nothing here is a vendor pitch, and where we do not know something we say so rather than filling the gap.
Data centres and colocation
Choosing a facility, reading what its public record does and does not prove, and what the certifications are worth once you are in a contract.
Carrier diversity and certification signals in datacenter risk assessment
How to interpret carrier presence records and certification signals when assessing datacenter infrastructure risk for insurance underwriting, compliance, and vendor qualification.
Carrier-neutral vs single-carrier colocation: what matters for your use case
When carrier neutrality changes the colocation answer, when a single-carrier operator is the right pick, and how to read an operator's network footprint before signing.
How to build a datacenter shortlist before your first vendor call
A practical process for procurement teams: build a defensible datacenter shortlist from public data before engaging vendors, so the first call is a qualification conversation rather than a discovery one.
How to choose a colocation provider
A practical checklist for choosing colocation: connectivity density, location and latency, power and cooling, certifications, operator credibility, and the questions to ask before you sign.
How to choose a colocation provider: a procurement checklist
A practical checklist for colocation procurement covering connectivity density, location and latency, power and cooling, certifications, operator credibility, and contract terms.
How to verify datacenter operators during acquisition due diligence
A practical framework for M&A teams: confirm operator identity against network evidence, trace ownership chains through corporate registries, and flag sanctions exposure before signing.
Internet exchange colocation: what an IX is and why it matters
An internet exchange (IX) lets networks peer directly inside a facility, cutting latency and transit cost. Here is why colocating where an IX is on-net is a connectivity advantage.
Reading a datacenter facility record: what the public data actually tells you
A field-by-field guide to reading a datacenter facility record: which data points come from independent sources, which come from operator self-reporting, and what each actually tells a procurement buyer.
Sanctions risk in datacenter ownership chains: what acquirers miss
How sanctions exposure reaches datacenter acquisitions through subsidiary structures, beneficial ownership, and operator arrangements that standard diligence checklists routinely miss.
Tier III vs Tier IV data centers explained
Tier III is concurrently maintainable; Tier IV is fault tolerant. The difference is redundancy and uptime, and most buyers do not need Tier IV. Here is how to choose.
What SOC 2 and ISO 27001 mean for your data center contract
How procurement teams should read SOC 2 Type II and ISO 27001 certifications in colocation procurement, what each report actually covers, and the contract clauses that operationalize them.
What is a carrier-neutral data center?
A carrier-neutral data center lets multiple networks and carriers interconnect, so you are not locked to one provider. Here is why it matters for latency, redundancy, and price.
What underwriters look for when assessing datacenter infrastructure
The signals cyber and infrastructure underwriters use to evaluate datacenter risk: operator verification, certifications, carrier diversity, physical resilience, and ownership chain exposure.
Section index: /colo/guides · /datacenters/guides
GPU compute
Telling direct operators apart from marketplaces and resellers, and working out what an hour of compute really costs.
Direct, marketplace or reseller: which GPU cloud type fits your workload
How to tell direct GPU operators, marketplaces and resellers apart by looking at network, capacity shape and contract structure and which sourcing model fits which kind of workload.
How to choose a GPU cloud provider
A short, opinionated framework for shortlisting GPU cloud providers without getting lost in spec sheets or per-hour numbers that hide the real cost.
What does an H100 actually cost
Why H100 list prices range from under two dollars an hour to well above ten, what the headline number leaves out and how to compute a realistic monthly bill before signing.
Section index: /gpu/guides
Carrier and bandwidth
Shortlisting transit providers, and the difference between peering and transit when you are the one paying for it.
BGP peering vs transit: what infrastructure buyers need to know
Peering and transit are two different ways to connect networks. This guide explains when each is the right choice, what settlement-free peering actually requires, and how internet exchanges fit in.
How to choose a bandwidth provider for low-latency applications
Latency requirements change which carrier criteria matter most. This guide covers geographic routing, exchange presence geography, latency SLAs, and how to test providers before committing.
How to evaluate an IP transit provider
A practical framework for shortlisting IP transit providers: network reach, exchange presence, SLA terms, pricing structure, and the questions to ask before signing.
Section index: /bandwidth/guides
Managed services
Selecting a managed IT partner, writing an RFP that returns comparable answers, and reading the certifications properly.
How to choose a managed service provider
The right MSP is an operating partner rather than a bundle of tools, so start with responsibility, evidence and escalation paths.
How to read MSP certifications
Certification badges are useful shortlist signals once a buyer checks scope, recency and the entity actually named on the evidence.
MSP RFP checklist
A useful RFP produces comparable commitments, so ask for ownership, limits and proof instead of broad capability statements.
Section index: /msp/guides · /msp/learn
Research notes
Findings taken from the verified directory itself, written up as evidence rather than opinion.
The most connected network in our IP transit set is the thirteenth biggest
Cogent announces 37 million addresses and appears at zero exchange points. Equinix announces a twentieth of that and sits at 78. Size and reach are not the same purchase.
We can prove a provider runs its own network but not that it is the right one for you
Most directory badges mean a company filled in a form. This one means we checked the routing. Here is exactly where that evidence stops.
Section index: /research