A data centre runs continuously, which means somebody is responsible for it continuously. How that responsibility is covered is one of the larger recurring costs in an operating budget and one of the least examined parts of a design.
Start from what actually has to be covered
Not every facility needs people on site at three in the morning. What every facility needs is a defined answer to three questions: who notices, who decides, and who can physically touch the equipment.
Monitoring systems answer the first. Escalation procedures answer the second. Only presence answers the third, and the cost of presence is what staffing models are really arguing about.
A site where a failure is detected remotely, diagnosed remotely, and resolved by a technician arriving within an hour may not need overnight presence at all. A site where the nearest competent person is three hours away has a different answer, regardless of how good the monitoring is.
The common models
Business hours plus on-call
Staff present during the working day, an on-call rota outside it. The cheapest model and entirely reasonable for smaller sites with good monitoring and a nearby team. The exposure is response time at the worst hour, and on-call rotas degrade if they are called often.
Extended coverage
Two shifts covering roughly sixteen hours, with on-call overnight. A middle position that puts people on site for the hours when most work happens without funding a full night shift.
Continuous 24/7
Four crews on a rotating pattern, commonly twelve-hour shifts, so one crew is always present. This is what large facilities and most colocation providers run. It is expensive, and the arithmetic surprises people: covering one seat continuously takes more than four full-time people once leave, training and sickness are accounted for.
Remote operations with local hands
A central operations centre watches several sites, with a small local presence or contracted technicians at each. This is how multi-site operators keep expert attention affordable, and it works well when the local hands have clear, bounded procedures.
The roles behind the shift pattern
Coverage is not one job. A functioning site needs several capabilities present or reachable.
- Operations technicians, who watch the systems, run rounds and handle first response.
- Electrical and mechanical specialists, who work on the power and cooling plant. Often shared across sites or contracted.
- Remote hands, who perform physical tasks on customer equipment: reseating, swapping, cabling, power cycling.
- Security, which is a separate function with separate requirements.
- A shift lead, who is authorised to make a decision at two in the morning without waking three people first.
That last one is the role most often left undefined. An escalation path that requires consensus is not an escalation path.
Where staffing models fail
Handover. Most things that go wrong across a shift boundary go wrong because a known issue was not passed on. A written handover, covering open issues, work in progress and anything deferred, is the cheapest reliability measure available to an operator.
Alarm fatigue. A team that receives hundreds of low-value alerts stops reading them, and then misses the one that mattered. This is a monitoring design problem that presents as a staffing problem.
Single points of knowledge. When one person is the only one who understands the chiller sequence, the site is exposed every time that person is on leave. Documented procedures and deliberate cross-training are what prevent it.
Night shift decay. Overnight crews that never see daylight management drift from procedure. Rotating people through shifts, rather than permanently assigning them, costs some convenience and buys consistency.
What this means for a colocation customer
If you are buying space rather than running it, the staffing model is a service question and it is worth asking directly.
- Is there someone on site at all hours, or is overnight covered by on-call?
- What is the committed response time for remote hands, and is it a commitment or a target?
- Is remote hands included, bundled into a monthly allowance, or billed per incident?
- Who is authorised to act on your equipment without waiting for you?
- How are shift handovers recorded, and will an issue raised at 6pm still be known about at 6am?
Our colocation contract checklist covers how these commitments should appear in an agreement rather than in a conversation.
Security staffing is a separate problem
Physical security is often folded into the same conversation and should not be. The skills differ, the shift pattern differs, and the failure modes differ.
A security officer watching a screen is not an operations technician and should not be relied on to interpret a cooling alarm. Equally, a technician escorting a visitor is not doing the job they were hired for. Sites that blur the two end up doing both badly at the moments when both matter.
Where budgets are tight, the useful compromise is clear: keep the roles separate during the hours when customer work happens, and combine them only where procedures are explicit about what the combined role is and is not authorised to do.
What good looks like on paper
A staffing model that works tends to have the same artefacts, and their absence is a reliable warning sign.
- A written shift roster that shows coverage, not just names.
- A standing handover format, used every shift, kept where the next crew will read it.
- Named escalation with authority levels attached, so a decision does not wait on a committee.
- Procedures for the half-dozen events most likely to happen, written for someone who has not seen that event before.
- A training record showing who is competent on what, and therefore where the single points of knowledge are.
None of it is exotic. It is the difference between a team that handles an incident and a team that discovers, during one, that nobody wrote anything down.
The practical point
Staffing is usually treated as a cost to minimise. It is more useful to treat it as the mechanism that converts monitoring into action, because a facility that detects a problem and cannot act on it for four hours has bought instrumentation and not resilience.
Rax Data and Energy staffs its facilities continuously and defines remote hands response in the agreement rather than leaving it to goodwill. If you want the operational commitments spelled out before you sign, get in touch.
Frequently Asked Questions
Does a data centre need staff on site 24/7?
Not always. What every site needs is a defined answer to who notices a problem, who decides, and who can physically touch the equipment. Where monitoring is good and a technician can arrive quickly, overnight presence may not be required.
How many people does continuous coverage actually take?
More than four. Covering a single seat around the clock needs four crews on a rotating pattern, and once leave, training and sickness are accounted for it takes more than four full-time staff per seat.
What is the most common operational failure in a staffed site?
Handover. Most problems that cross a shift boundary do so because a known issue was not passed on. A written handover covering open issues, work in progress and anything deferred is the cheapest reliability measure available.
What should a colocation customer ask about staffing?
Whether someone is on site at all hours or overnight is on-call, the committed remote hands response time and whether it is a commitment or a target, whether remote hands is included or billed per incident, and how shift handovers are recorded.
Operational Commitments in the Agreement, Not the Conversation
Rax staffs its facilities continuously and defines remote hands response in the contract. Talk to our team about what that means for your deployment.
Contact Rax Data & Energy