Data Center & Server Infrastructure

• Server inlet/outlet temperature (per-rack, ASHRAE thermal guidelines)
• Rack-level PDU: per-outlet current, voltage, power factor, circuit breaker state
• CRAC/CRAH unit performance: supply/return air temperature differential, airflow rate
• UPS: state of charge, estimated run-time remaining, battery internal resistance, cell voltage
• Generator: fuel level, load capacity, last test result, transfer switch state
• Raised floor: differential pressure (airflow management optimization)
• Hot aisle / cold aisle: temperature gradient mapping (thermal mapping array)
• Server IPMI/BMC telemetry: CPU temperature, fan speeds, DIMM error count, disk S.M.A.R.T.
• Network room: relative humidity and dewpoint (condensation risk)
• CRAH compressor: suction/discharge pressure, refrigerant state, compressor current

• PUE (Power Usage Effectiveness) trending → identify efficiency loss before it becomes

significant cost

• Thermal runaway prediction: hot spots developing in rack or cold aisle before

causing server crashes

• UPS battery degradation curve fitting → predict replacement need 6–12 months ahead
• Cooling capacity headroom vs actual load → predict when cooling becomes insufficient

for planned capacity growth

• Power chain topology analysis: identify single points of failure in PDU/UPS/ATS paths
• Correlated hardware fault prediction: disk S.M.A.R.T. + temperature history +

DIMM error rate → failure probability scoring

• Hot spot forming → Increase CRAC output in that zone, redistribute workload to cooler

racks, alert capacity planning

• UPS battery approaching end-of-life → Schedule replacement, alert facilities, reduce

load-to-backup-time commitment

• Generator test failure → Escalate to facilities, schedule maintenance, increase

inspection cadence until resolved

• PUE degrading → Identify root cause (clogged filters, CRAC setpoint drift, blanking

panel missing), alert facilities

• Disk predictive failure → Live-migrate workload, schedule disk replacement before

unplanned downtime

• PUE improvement year-over-year (cost and sustainability metric)
• UPS-dependent incident rate reduction
• Generator reliability score (percentage of successful test starts — target 99%+)
• Hardware replacement cost reduction via predictive vs emergency replacement

• Colocation data centers (Equinix, Digital Realty, CyrusOne model)
• Enterprise on-premise data centers
• Hyperscale cloud (edge and regional facilities)
• Telecom central offices and carrier hotels
• Financial services (trading floor infrastructure, low-latency colocation)
• Government and military data centers (FISMA, IL4/IL5 classified environments)
• Edge computing nodes (distributed, often unattended, limited cooling headroom)