Data center maintenance management and PM program field guide

What a data center PM program is, and why deferring is borrowing

A data center maintenance program is the scheduled and condition-based work that keeps the critical power and cooling plant reliable: the UPS and its batteries, the generators, the switchgear and distribution, the CRAC and CRAH units, the chillers and towers, and the CDUs feeding liquid-cooled racks. It runs on three modes at once, preventive on a calendar, predictive on the condition of the gear, and reactive when something breaks, and the whole point is to do that work without ever dropping the load the building exists to carry.

Here is the part nobody wants on the budget call. Deferring maintenance on critical gear is borrowing against the next outage. A skipped PM, a battery jar past its life, a generator that has not seen a real load in a year, none of it shows on the dashboard. The plant looks healthy right up until the utility blinks and the redundancy you paid for is not there. The redundancy is only real if it is maintained.

This guide covers the program: the maintenance types, the PM by equipment, the OEM intervals, the approved procedures, and the CMMS that tracks it. Two pieces have their own guides. The battery, which is the part most likely to fail quietly, is covered in the UPS battery maintenance and testing guide. The baseline the program inherits comes from the commissioning levels guide. This is the program that carries that baseline forward for the life of the building.

Neglected gear fails at the worst time

Critical equipment does not fail at random. It fails when it is run hard and neglected, and it tends to fail at the exact moment it is finally called: a utility loss, a transfer, a hot afternoon at full IT load. That is the cruel arithmetic of a data center. The gear sits in standby or in a redundant leg for months, and the one time it has to perform is the one time you find out the maintenance was skipped.

The mechanism is the latent fault. A loose lug, a worn contactor, a battery string down a third of its capacity, a clogged coil, a generator with stale fuel, each one is a failure already present in the plant, waiting for the conditions that expose it. Preventive and predictive maintenance exist to find the latent fault on a scheduled afternoon, on your terms, instead of at 2 a.m. with customers on the line.

Run the redundancy logic and it gets sharper. A site is built N+1 or 2N so a component can fail and the load rides through. But if the redundant unit is the one carrying a deferred PM, a single failure becomes an outage, because the backup was never ready. Maintained redundancy is the only redundancy. The drawing says 2N. The maintenance program is what makes that true.

What is the difference between preventive, predictive, and reactive maintenance?

There are three ways to maintain a piece of equipment, and a mature program uses all three deliberately. Preventive maintenance is scheduled on a calendar or runtime interval, change the filter, torque the connections, service the chiller, whether or not anything is wrong yet. Predictive maintenance is condition-based, you watch the actual health of the gear with thermography, vibration, oil analysis, and battery monitoring, and you act when the data says the part is heading for failure. Reactive maintenance is run-to-failure, you fix it after it breaks.

The lean is clear. Reactive-only on critical gear is the most expensive and dangerous mode, because the failure picks the timing and the timing is always bad. Preventive is the floor every program has to hold. Predictive is the upgrade that catches the failures a calendar misses, the bearing going bad between services, the connection heating up under load.

The maturity arc of a critical-facilities program is moving work from reactive toward predictive without ever abandoning preventive. You keep the OEM preventive intervals as the baseline, layer predictive monitoring on the gear that justifies it, and shrink the reactive bucket to genuine surprises. Reactive never goes to zero. The goal is that it stops being your strategy and becomes the exception.

Preventive maintenance: the scheduled floor

Preventive maintenance is the scheduled work that keeps gear inside its design envelope before anything goes wrong. It is the foundation of the program because it is predictable, plannable, and tied to intervals the manufacturer sets. Filters get changed, connections get torqued and thermally checked, fluids get sampled and changed, breakers get exercised, belts and bearings get serviced, all on a clock or a runtime counter rather than on a failure.

Build the PM schedule off the OEM maintenance manuals first, then adjust for how hard the site runs. A unit at 60 percent load in a clean room ages differently than one pinned near capacity in a dusty space, and the interval should reflect the duty. The manufacturer's recommended interval is the starting basis, and the project specification or service contract may tighten it.

The discipline that separates a real program from a binder is completing the PM on time and recording what was found. A PM marked complete with no as-found readings is a checkbox, not maintenance. The value is in the trend, the torque that had backed off, the filter loaded heavier than last quarter, the coil temperature creeping up. That is preventive work feeding the predictive picture.

Predictive maintenance: catch it before it fails

Predictive maintenance reads the condition of the gear and acts before the failure, instead of on a fixed date. It is the upgrade that catches what a calendar cannot, the fault that develops between scheduled services. On critical power and cooling equipment the payoff is direct: you find the problem while the unit is still running and the redundant leg is still intact, not after it has dropped.

Four techniques carry most of the load. Infrared thermography finds loose or corroded connections and overloaded conductors by the heat they make, which is why an annual IR scan of switchgear, transfer switches, breakers, and distribution catches problems a visual check misses. Vibration analysis predicts bearing and rotor wear on chillers, pumps, fans, and motors before they seize. Oil analysis on chiller compressors and generator engines reads wear metals and contamination from inside the machine. Battery monitoring, the ohmic and capacity trending covered in the UPS battery guide, warns that a string is going long before a discharge proves it.

Predictive does not replace preventive, it sharpens it. The IR scan and the vibration route tell you which PM to pull forward and which interval has slack. Tie the readings to a trend in the CMMS, because a single thermal image is a snapshot, and the trend across quarters is what actually predicts the failure.

The critical equipment and what its PM covers

A data center PM program touches every piece of gear in the power and cooling path, and each has its own tasks and intervals. The list below is the working scope of the program. The intervals are typical industry practice and a place to start, not a substitute for the OEM manual, the NFPA and NETA basis where it applies, and the site's own duty and spec, all of which control the actual schedule.

The electrical side runs from the UPS and its batteries through the switchgear, breakers, transfer switches, PDUs, and transformers, with the generators standing by behind them. The mechanical side runs from the chillers and cooling towers through the CRAC and CRAH units, pumps, and, increasingly, the CDUs that feed liquid-cooled racks. Both sides have to be maintained together, because the cooling plant is as critical as the power plant the day the room heats up.

UPS and batteries: the battery is the weak link

The UPS is the ride-through that carries the load across the gap between utility loss and the generator coming up to speed, and its PM is real: capacitor and fan inspection, an IR scan of the power connections, calibration, and firmware on the OEM schedule. Capacitors and fans are wear items with a finite life, and a UPS that has run years past its capacitor service is a failure waiting on a transfer.

But the part most likely to let you down is the battery, and it fails in a way nothing else does, silently. A string on float reads full voltage and gives every sign of health while it quietly loses capacity to heat and age. Float voltage tells you almost nothing about runtime. Only ohmic trending and a capacity discharge test tell you whether the autonomy the owner paid for is actually there.

Because the battery is its own discipline, the full treatment, VRLA versus flooded versus lithium-ion, the capacity test, impedance baselining, float and temperature control, and the replacement criteria, lives in the UPS battery maintenance and testing guide. For the program, the rule is simple: the battery gets its own PM line, its own trend, and its own budget, and you do not let it ride on a voltage reading.

Generators: the standby that has to start

The generator is standby power, which means it does nothing for months and then has to start, take load, and run for as long as the outage lasts. Its PM exists to guarantee that one event. The work splits into the regular exercise, the periodic load test, and the fuel program, and the basis for the schedule on emergency and standby systems is NFPA 110 plus the engine OEM's intervals.

Under NFPA 110, the emergency power supply system is commonly inspected weekly and exercised under load at least monthly, with the monthly run long enough for the engine to reach operating temperature, often cited as at least 30 minutes. A no-load or lightly loaded run is not enough, because an engine that never sees real load wet-stacks and glazes its cylinders. That is what the load bank is for. Load bank testing is commonly an annual requirement under NFPA 110, with the longer-interval full-capacity test on the longer cycle. Confirm the current edition and the adopted requirements.

The fuel is the quiet killer. Diesel degrades, takes on water, and grows microbial contamination sitting in a tank for a year, so fuel sampling and polishing belong in the program. The generator that cranks fine on the monthly exercise can still die twenty minutes into a real outage on bad fuel. Test the fuel, not just the start.

Switchgear and breakers: the power chain

The switchgear and the breakers carry and protect the power path, and their PM is where electrical reliability is won or lost between the UPS and the load. The work is connection integrity, breaker function, and protective-device testing, and the acceptance and maintenance basis for the electrical gear is commonly NETA, with NFPA 70B, retitled a standard with mandatory requirements as of its 2023 edition, as the basis for the maintenance program itself.

Infrared thermography is the highest-value predictive check on the power chain. Loose and corroded connections, overloaded conductors, and failing components all make heat before they fail, and an IR scan under load finds them while a visual inspection sees nothing. An annual scan of the switchgear, transfer switches, breakers, and distribution is common practice, and the connections it flags get torqued to the manufacturer's value and rechecked.

Breakers themselves are mechanical devices that age in place. A breaker that has sat closed for years may not open when it is finally asked to, so exercising and testing breakers on a maintenance interval, including the protective relays and trip units, is part of the program. The interval and the test procedure follow NETA and the OEM. Live electrical work is where the safety stack matters most, and none of it happens outside an approved procedure with the right PPE and arc-flash boundary.

Cooling gear: the plant that keeps the room alive

Cooling is as critical as power, because a room that loses cooling at full IT load overheats in minutes, not hours. The cooling PM runs across the CRAC and CRAH units, the chillers, the towers, and the CDUs, and the failure modes are mostly about airflow, water, and refrigerant.

On the air side, CRAC and CRAH units live and die by filters and coils. A loaded filter or a fouled coil starves airflow and drives the unit to work harder for less cooling, so filter changes on a monthly-to-quarterly cadence, coil cleaning, belt and bearing service, condensate management, and a refrigerant check are the core tasks. On the water side, chillers carry the heavy PM: oil and vibration analysis, condenser-tube cleaning, refrigerant and controls. Condenser scale matters more than people expect, because even a thin scale layer cuts chiller efficiency measurably. Cooling towers need continuous water treatment to control scale, corrosion, and biological growth, plus basin and fill cleaning.

CDUs are the newer piece, driven by liquid-cooled AI and HPC racks. Their PM is coolant quality, filtration, pump service, and leak detection, with filtration commonly in the fine micron range to protect the cold plates. Many CDUs carry dual filters so a filter can be changed while the unit keeps running. Coolant quality is the item that gets ignored until a cold plate fouls, so it belongs on the schedule with a real sample, not a glance.

Should you follow the OEM maintenance intervals?

Yes. The OEM maintenance intervals are the basis the PM program is built on, for two reasons. The manufacturer knows the wear items and the failure modes of its own gear better than anyone, and the warranty and any extended service agreement are usually written against those intervals. Miss them and you can void coverage on a six-figure piece of equipment at the moment you most need the manufacturer to stand behind it.

Treat the OEM interval as the floor, then adjust for duty. A unit running hot, dirty, or near capacity earns a tighter interval than the manual's nominal recommendation, which is typically written for moderate conditions. The project specification, the service contract, and the standards that apply, NFPA 110 for the generator, NETA and NFPA 70B for the electrical gear, IEEE practice for the batteries, can all tighten the schedule beyond the manual. Where they conflict, the stricter requirement governs.

What you do not do is invent intervals to fit a budget. If the manual says annual and the budget says every two years, that gap is deferred maintenance with a different name, and the warranty language will say so when a claim comes in.

Can you maintain a data center without downtime?

Yes, and a properly designed site is built specifically so you can. The feature that allows it is concurrent maintainability, the design principle behind a Tier III facility, where every component in the power and cooling path can be maintained, repaired, or replaced without taking the IT load offline. It works through redundant distribution paths and redundant units, so the load shifts to one leg while you work on the other.

The mechanics on the floor are straightforward and unforgiving. You isolate the unit you are servicing, confirm the redundant leg is carrying the load and healthy, do the work, and bring the unit back, all without the critical bus ever seeing an interruption. An N+1 cooling plant lets you pull one CRAH for service while the others hold the room. A 2N power path lets you maintain one side dead while the other carries everything.

The discipline is that no single maintenance activity is allowed to put the site in a single point of failure it cannot ride. The moment you take the redundant unit offline for PM, the running unit is now N, with no backup, so you keep that window short and you do not stack two activities that both lean on the same surviving path. The redundancy and tier design is its own subject. For maintenance, the rule is: maintain the leg that is offline, never the leg carrying the load, and never both at once.

Every task runs under an approved MOP

No work touches critical gear in a live data center without an approved Method of Procedure. A MOP is the step-by-step script for a task that changes the state of any critical component, written and approved before anyone lifts a tool. It is the single most important control in the program, because the largest cause of data center outages is not equipment, it is human error, and a MOP is how you take the error out of the work.

A real MOP carries more than the steps. It has the risk assessment, the back-out or rollback plan if the work goes wrong, the verification that redundancy is in place before the activity starts, and the named approvals. Before a tech isolates a UPS module or racks out a breaker, the MOP has already confirmed the load is on the other leg and spelled out exactly how to reverse course if something does not behave.

The related controls are the SOP for routine operation and the EOP for emergencies, but the MOP is what governs planned maintenance specifically. Run a PM on critical gear without one and you are improvising on the live plant. That is the day a wrong breaker gets opened. Write the MOP, get it approved, and follow it step by step, including the rollback.

The CMMS: where the program actually lives

A maintenance program without a system to run it is a stack of good intentions that slips the first busy month. The CMMS, the computerized maintenance management system, is the work-order and scheduling tool that holds the program together. It schedules the PMs, generates the work orders, captures what was done and what was found, tracks the parts, and keeps the equipment history that turns isolated services into a trend.

The reason a critical-facilities program cannot run on a spreadsheet is accountability. The CMMS is what tells you, on any given day, which PMs are due, which are overdue, what the backlog looks like, and whether the gear is trending toward trouble. Uptime Institute's M&O criteria treat a maintenance management system that tracks status and trends activity as a requirement of an effective program, precisely because it is the only way to keep deferred maintenance visible instead of invisible.

Whatever runs the program has to be usable by the tech standing in front of the gear, not just the manager at a desk. A field tool like FieldOS that lets the technician pull the work order, follow the MOP steps, record the as-found and as-left readings, and capture photos at the equipment closes the loop that a back-office system leaves open. The history is only as good as what the person doing the work actually captured.

Tracking PM compliance

PM compliance is the metric that tells you whether the program is real or aspirational. It is the percentage of scheduled preventive maintenance completed on time within its window, and it is the single number a critical-facilities manager should be able to recite. A program at 98 percent on-time is maintaining its gear. A program at 70 percent is carrying a hidden backlog of deferred work, and the gap is latent risk accumulating on the floor.

On-time matters as much as done. A PM completed three months late on a unit that failed in month two is not compliance, it is a near miss the metric should have flagged. Track completed-on-time against scheduled, watch the overdue count and the backlog age, and treat a rising backlog as the early warning it is.

Set the target with the gear's criticality in mind. The redundant generator and the UPS battery do not get the same tolerance for a slipped PM as a comfort fan in an office. Many programs hold critical-equipment PM to a near-100 percent on-time standard and accept more slack on non-critical items. The exact threshold is the operator's call, set against the site's risk posture and any contractual or M&O commitment.

What happens if you defer critical maintenance?

Deferring critical maintenance does not save money, it moves the cost to a worse time and adds interest. Every deferred PM is a latent fault left in the plant, and Uptime Institute's M&O guidance is blunt that any deferred maintenance activity becomes a risk to the data center. The risk does not sit still. A skipped service becomes a worn part, a worn part becomes a failure, and a failure on the redundant leg becomes the outage the redundancy was supposed to prevent.

Deferral almost always comes from budget pressure, and it is seductive because nothing breaks the day you skip the PM. The plant looks fine. The dashboard is green. That is exactly the trap, because the gear that is most deferrable in the budget meeting, the standby generator, the redundant chiller, the battery string, is the gear whose failure you will not see until the day it is called.

The discipline is to protect critical PM as non-negotiable and let deferral, if it must happen, fall on genuinely non-critical work, scored by criticality and risk. Reliability-centered methods exist to make that call defensibly, removing low-risk items from the schedule on purpose rather than letting high-risk items slip by neglect. Defer the critical and you are borrowing against the next outage at a rate you will not like when the bill comes.

Critical spares and long-lead parts

A maintenance program is only as fast as the part it needs, and the failure with no spare on the shelf is the one that turns a quick repair into a multi-week outage. Critical spares are the parts whose absence extends a failure into a real risk, and they belong in inventory before they are needed, not ordered after the gear is down.

The parts that hurt are the long-lead items. Switchgear components, breakers, large transformers, chiller compressors, and UPS power modules can carry lead times of months, and that lead time has stretched as the industry has grown. A redundant site can ride a failure on the surviving leg, but only for as long as the failed unit stays repairable, and if the part is twelve weeks out, the site runs without redundancy for twelve weeks. That is a long time to be one failure from an outage.

Stock the spares the criticality and the lead time justify, hold them to the OEM's storage requirements, and rotate the ones with a shelf life. The supply and procurement side is its own discipline, but the maintenance program owns the question of which parts must be on hand, because the program is what knows the failure modes and the gear that has no tolerance for a wait.

OEM service contracts and managing the vendor

Some of the PM on critical gear is specialized enough that the OEM or a qualified service vendor does it, not the in-house team. UPS firmware and module service, chiller major overhauls, generator engine work, and switchgear testing often fall under an OEM or third-party service contract, and managing those contracts is part of running the program.

The terms that matter are scope, interval, and response time. The contract should spell out exactly which PMs the vendor performs and on what schedule, and the emergency response time, how fast a technician is on site when critical gear fails, because a 24-hour response on a UPS with no redundancy left is not the same product as a 4-hour response. For the gear that the warranty ties to OEM service, the contract is also what keeps the coverage alive.

The operator still owns the program. A vendor doing the work does not transfer the responsibility for verifying it got done, that the as-found and as-left data came back, and that the service report goes into the CMMS history like any other PM. The broader operations and vendor-management practice is its own subject. For maintenance, the rule is: contract the specialized work, but hold the vendor to the same record and the same intervals you hold the in-house team.

The maintenance-versus-uptime tension

There is a real fear in this work, and it is worth naming. Touching live critical gear feels more dangerous than leaving it alone, because the act of maintenance, isolating a unit, racking a breaker, transferring a load, is itself a moment where something can go wrong. Plenty of outages have been caused by the maintenance, not the lack of it. So the instinct to leave the running plant alone is not irrational. It is just wrong.

Not maintaining is the bigger risk, and the difference is timing and control. The maintenance failure happens on a scheduled afternoon, under a MOP, with the redundant leg carrying the load and a rollback plan ready. The neglect failure happens whenever the latent fault decides, with no plan and no margin. One you control. The other controls you.

The resolution is not to maintain less, it is to maintain better: every task under an approved MOP, redundancy verified before the work starts, the procedure followed step by step, and a back-out ready. That discipline is what makes touching the live plant safer than fearing it. The site that is afraid to maintain its gear is the site quietly accumulating the failures it is afraid of.

Documentation: as-found, as-left, and the trend

The maintenance record is the proof the work happened and the data that predicts the next failure, and a program that does not document well is flying blind no matter how much work it does. The two readings that matter on every task are as-found and as-left, the condition the technician encountered and the condition they left. As-found is where the trend lives. A connection found loose, a filter found heavily loaded, a battery jar found drifting, that is the early warning, and it only exists if someone wrote it down.

The records build three things at once. They prove PM compliance and warranty adherence when an auditor or an OEM asks. They build the equipment history that shows whether a unit is stable or sliding. And they give the trend that turns preventive work into predictive insight, because the value of an IR scan or a vibration reading is the change across quarters, not the single number.

This is where a field tool earns its place. A technician using FieldOS to capture the as-found and as-left values, the photos, and the completed MOP steps at the gear, instead of filling out a form from memory at the end of the shift, produces a record that is actually trustworthy. The trend is only as good as the field capture behind it, and the field capture is only as good as how easy it is to do standing in front of the equipment.

Maturing into a reliability program

The most mature version of a maintenance program stops asking only what the manual says and starts asking what each piece of gear actually does and how it actually fails. That is reliability-centered maintenance, RCM, a method that sets the maintenance strategy for each asset based on its criticality, its function, and its failure modes rather than a one-size calendar. It came out of aerospace and has been adopted across mission-critical facilities for the same reason: not every part deserves the same maintenance.

RCM lets you spend the maintenance budget where the risk is. The critical, hard-to-replace, no-redundancy asset gets aggressive preventive and predictive coverage. The low-risk, redundant, easily-replaced item may be safe to run closer to failure, which is how RCM defensibly trims a backlog instead of letting it rot. The decision is criticality and risk, scored, not guessed.

This is a maturity, not a starting point. A site gets there by running solid preventive and predictive work first, building the failure history in the CMMS, and then using that history to optimize the program to the real failure modes. Done well it raises reliability and lowers cost at the same time. Done as a label on the same old schedule it changes nothing. The data has to be real.

Commissioning sets the baseline the program carries

A maintenance program does not start from zero, it starts from the numbers commissioning recorded when the plant was new. Commissioning proves the systems work and survive failures before any IT load arrives, and it captures the as-built, as-tested baseline: the clean torque values, the load bank results, the battery capacity at install, the airflow and the setpoints. That baseline is the reference every future PM trends against.

The handoff is where programs succeed or fail. The commissioning record, the levels and the integrated systems test, lives in the data center commissioning levels guide, and the operating team should inherit it as the starting condition of the gear. When a battery string is at 100 percent of rated capacity at commissioning, the maintenance trend that watches it slide toward the 80 percent replacement threshold only makes sense against that baseline number.

Maintenance is commissioning continued for the life of the building. The same discipline that proved the plant on day one, witnessed tests, recorded data, failure scenarios, is what the PM program carries forward at every interval. Lose the commissioning data in the handoff and the program is trending against nothing for the first several years, which is exactly the window where early-life failures show up.

Field checklist

What to document

The PM record has to answer, months later, what was done, what condition the gear was in, and whether it is trending toward trouble. Capture it at the equipment, not after, and tie every entry to the asset so the history accumulates into a trend instead of scattering into loose reports.

At minimum, record the equipment and asset ID, the PM or task performed, the MOP followed and its approval, the as-found and as-left readings, any parts replaced, the predictive data captured, who did the work, and the date against the scheduled window. If a finding triggers a follow-up, that becomes its own tracked item, not a note that dies in a report.

Common mistakes

• Deferring critical preventive maintenance under budget pressure because nothing breaks the day you skip it.
• Running reactive-only on critical gear and calling it a strategy instead of a string of near misses.
• Doing maintenance on live critical gear without an approved MOP and a rollback plan.
• Running the program on spreadsheets or memory with no CMMS, so PMs slip and the backlog goes invisible.
• Ignoring battery, IR, vibration, or oil warning signs because the gear has not failed yet.
• Carrying no critical spares, so a failure that should be a quick repair becomes a multi-week outage.
• Taking the redundant unit offline for PM and forgetting the running unit is now N with no backup.
• Marking a PM complete with no as-found data, leaving the trend with nothing to trend.

Standards and references

The maintenance program leans on the OEM and several standards bodies, and each governs a different part. The OEM maintenance manuals and any service or warranty agreement are the primary basis for intervals and tasks on each piece of gear, because the manufacturer defines the wear items and the warranty is written against them. Where a standard applies, it sits alongside the manual and often tightens it.

For the generator and the emergency power supply system, NFPA 110 governs inspection, exercise, and load testing, with the engine OEM adding its own service. For the electrical gear, NETA gives the acceptance and maintenance testing basis for switchgear, breakers, and protective devices, and NFPA 70B, retitled a standard with mandatory requirements in its 2023 edition, sets the electrical maintenance program, with NFPA 70E governing the safety side of working on or near live equipment. For batteries, IEEE maintenance and testing practice plus the battery OEM control, as covered in the UPS battery guide.

For the program itself, Uptime Institute's Management and Operations criteria set the framework for a maintenance management system, deferred-maintenance tracking, and procedures, and ASHRAE TC 9.9 governs the thermal envelope the cooling plant has to hold. Reliability-centered maintenance gives the method for optimizing the program to failure modes. The hierarchy is steady across all of it: maintain the redundancy or it is not real, run preventive plus predictive on the critical gear, and do every task without dropping the load under an approved MOP. The intervals, tests, and acceptance criteria are set by the OEM, the applicable standard, and the operator, and those control over any rule of thumb here.

Terms and definitions

The maintenance world has its own vocabulary, and the same idea shows up under different names across an OEM manual, a service contract, and an M&O audit.

The terms below are the working set for a critical-facilities PM program. Where a definition carries a threshold or a basis, that threshold follows the OEM, the standard, and the project spec, not the general usage here.

Preventive maintenance (PM)

Scheduled maintenance on a calendar or runtime interval, performed before failure, on the OEM basis

Predictive maintenance (PdM)

Condition-based maintenance triggered by measured data (IR, vibration, oil, ohmic) before failure

Reactive maintenance

Run-to-failure repair after the equipment breaks; the exception on critical gear, never the plan

PM program

The full scheduled and condition-based maintenance regime for the critical power and cooling plant

CMMS

Computerized maintenance management system; schedules PMs, holds work orders, parts, and equipment history

MOP

Method of Procedure; the approved step-by-step script, with risk and rollback, for changing a critical component's state

Concurrent maintainability

Design that lets any component be maintained without taking the IT load offline, via redundant paths

Deferred maintenance

Scheduled maintenance not completed on time; a latent risk that grows until corrected

PM compliance

Percentage of scheduled PM completed on time within its window; the program's core metric

RCM

Reliability-centered maintenance; sets each asset's strategy by criticality, function, and failure modes

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• How do you test a UPS battery?
• What is data center commissioning?
• What is a blanking panel in a server rack?
• Why does a data center need a formal incident process?
• Why is cable management important in a data center?