Digital infrastructure is the interconnected compute, storage, network, identity, platform, endpoint, facility, and external-service foundation used to deliver digital services. Operations is the continuing work of making those services reliable, secure, observable, recoverable, supportable, and cost-aware.

Map the service, not just the hardware

LayerQuestions
Compute/platformCapacity, scheduling, isolation, patching, lifecycle and failure domains
Network/edgeRouting, DNS, certificates, latency, segmentation, filtering and dependencies
Storage/dataDurability, consistency, encryption, backup, retention, deletion and recovery
IdentityAuthentication, authorization, privilege, machine identity and emergency access
OperationsTelemetry, objectives, alerts, change, incidents, runbooks, support and learning

Define service objectives

Specify availability, latency, correctness, freshness, durability, recovery time and point, security, support, capacity, and cost objectives from user needs. A redundancy diagram is not evidence of resilience; test real failure and recovery paths.

Operate through evidence

  • Maintain ownership, dependency, data-flow, configuration, asset, and lifecycle records.
  • Collect metrics, logs, traces, audit events, synthetic checks, and user-visible outcome signals.
  • Alert on actionable symptoms and exhaustion, not every metric change.
  • Use reviewed changes, progressive delivery, rollback, backups, and rehearsed incident/disaster recovery.
  • Review incidents without assuming a single root cause; track contributing conditions and corrective-action evidence.

Secure every administrative path

Use least privilege, phishing-resistant authentication where appropriate, network segmentation, hardened endpoints, protected secrets, vulnerability/patch management, encryption, supply-chain controls, and logging. Review network security and endpoint management.

Choose architecture by constraints

Cloud computing changes responsibility rather than eliminating operations. Microservices can improve independent change and scaling for some systems while increasing distributed failure, data, testing, and observability complexity. Prefer the simplest architecture that meets measured requirements.

Include people, cost, and lifecycle

Document on-call load, safe staffing, accessibility, skills, vendor concentration, licenses, energy/resource use, total cost, capacity headroom, obsolescence, and exit. Retire unused services, revoke access, preserve required records, and verify data deletion.

Historical article substantially updated September 4, 2026. Original publication date preserved.