Check: geographic location and latency, tier certification, power and cooling redundancy, carrier diversity, physical security, and compliance (ISO 27001, SOC 2, etc.). Visit or audit if the workload is critical.
Location and latency
- Geography: Choose a region close to users or to other systems you integrate with. Latency and legal (data residency) both matter.
- Carrier diversity: Multiple transit and peering options reduce single-point-of-failure. Check how many carriers and whether the DC is carrier-neutral.
- Redundancy: Diverse fiber paths and redundant uplinks. Ask about historical uptime and incident reports.
Tier and resilience
- Tier certification: Tier I–IV (e.g. Uptime Institute) indicates design redundancy (power, cooling, network). Tier-III is common for production; Tier-IV for highest availability.
- Power: N+1 or 2N UPS, backup generators, and tested failover. Ask about runtime and fuel contracts.
- Cooling: Redundant cooling (N+1); hot/cold aisle containment. Efficient cooling lowers PUE and cost.
Security and compliance
- Physical security: 24/7 guards, access control, CCTV, mantrap. Who can get to your cages or racks?
- Compliance: ISO 27001, SOC 2, PCI-DSS, or industry-specific. Ask for current certificates and scope. Data residency and audit rights matter for regulated workloads.
- Visit or audit: For critical workloads, schedule a site visit or third-party audit. Verify what you are told on paper.
Summary
Check location/latency, tier, power/cooling redundancy, carrier diversity, physical security, and compliance. Visit or audit when the workload is critical.
