EPI Data Centre Foundation study materials
EPI Data Centre Foundation study resources

The 500 kW signature the site cannot serve

EPI Data Centre Foundation · Put your knowledge to work in context.

All case studies

The 500 kW signature the site cannot serve

You are site manager of a 2.4 MW facility with 60 percent of contracted capacity sold. Engineering assessment shows 380 kW of genuine spare capacity in Hall C after N+1 and at the design ambient, with no additional cooling available until the third quarter of next year. A sales director wants to sign a five-year contract for 500 kW starting in six weeks with a 99.999 percent availability SLA. The client's IT estate is dual-corded at the server level, but their racks and PDUs are single-corded and they have asked to be served from a single row busway. The facility's own contracted availability commitment to existing tenants is 99.98 percent. The sales director says to sign now and 'fix capacity later', and mentions that your promotion review is in three weeks and that 'flexibility is what gets people promoted'.

1. Set out what you do in the next 48 hours, in order, and what you refuse to do.

Read the worked answer

Quantify first, because a disagreement about numbers is resolvable and a disagreement about attitude is not. The commitment is 500 kW against 380 kW of verified spare at design ambient, so the shortfall is 120 kW and it lands on cooling, and possibly on UPS and generator headroom depending on the hall's topology. Write that down as a capacity statement with its basis: design ambient, redundancy state, PUE, and diversity assumptions.

Then escalate in writing to the executive who owns facility risk, not to sales, with three options: reduce the contracted load to within 380 kW with a priced growth option, provision temporary cooling with its associated power and floor space, or sign with a contractually capped ramp and a delivery date tied to the Q3 cooling. Copy the capacity owner and, if the SLA is in scope, legal or commercial risk, so the credit exposure is visible.

What you refuse to do is sign, verbally or otherwise, an availability and capacity commitment you know the site cannot deliver. A signature transfers the risk permanently to operations and to you. Record the decision and the named acceptor in the risk register. If the promotion is used as leverage, that is itself a governance failure worth reporting through the ethics or whistleblowing channel.

2. Assess the 99.999 percent SLA request against the facility's actual capability and the client's own configuration.

Read the worked answer

99.999 percent permits about 5 minutes 15 seconds of unplanned downtime per year. Test that against what the facility can actually deliver.

The facility commits 99.98 percent to existing tenants, which is roughly 1 hour 45 minutes per year. Offering one tenant a tighter SLA than the infrastructure, and tighter than the commitment to everyone else, means the first planned maintenance event or component failure breaches the new contract and triggers service credits. The site also lacks demonstrated full fault tolerance, so the number is aspirational rather than engineered.

The client's own configuration makes it undeliverable regardless of the facility. Single-corded racks and PDUs on a single row busway mean a rack-level power path failure is an IT outage no matter how redundant the upstream plant is. Availability cannot exceed the weakest single point, and theirs sits inside their own rack.

The honest answer is an SLA at or near 99.98 percent, with a clear demarcation of responsibility at the rack PDU and a written statement that higher figures require the client to adopt dual-corded A/B distribution inside their own space.

3. Who should own this risk acceptance, what must be recorded, and what does professional competence require of you here?

Read the worked answer

Risk acceptance belongs with the executive accountable for the facility's operational and commercial outcomes, typically the COO or the accountable director, acting on a documented technical assessment. It does not belong with sales, and it does not belong with the site manager, who does not control the commercial decision and should not be asked to underwrite it.

The register entry must record the risk in plain terms (a 120 kW capacity shortfall against a signed commitment with no cooling until Q3), the quantified exposure including SLA credit liability, the mitigation and its date, the residual risk, the named acceptor, and an expiry that forces review. The technical assessment signed by the capacity manager and the site manager sits behind it, so the facts cannot be edited later.

Professional competence here is not technical knowledge; it is refusing to represent something you know to be untrue. A site manager who signs or endorses a capacity and availability claim that the engineering contradicts has misrepresented the facility, exposed every tenant in the hall, and compromised their own standing. The correct professional position is to state the facts, offer options, document the decision whoever makes it, and let the accountable executive own the consequence.