Skip to content
SITECH

Six hours of outage a day: sizing a server room that stays up

Energy

Courbe de disponibilité électrique entrecoupée de délestages, avec relais par onduleur, groupe et solaire.

Agence Ecofin reported load shedding averaging up to six hours a day in the country's main cities, with Brazzaville and Pointe-Noire at the front. At the same time, Congo enjoys roughly two thousand hours of sunshine a year, while fewer than half the population has reliable access to electricity. In other words: the problem is daily, and the hybrid answer is no longer a conviction, it is a calculation.

Almost nobody runs that calculation seriously. A "solid" UPS is bought, a generator is added, and on the third outage of the day the room goes down anyway. Here is the method we apply on a site survey.

Step 1 — Measure the real load, not the rated load

The sum of the ratings printed on your server power supplies bears no relation to what they draw. A server with a 750 W supply often pulls 200 to 300 in normal use. Conversely, the total systematically forgets: room cooling — often the largest item — PoE-powered switches and access points, the storage array, emergency lighting and access control.

The survey runs over a week, with a clamp meter or the rack's own sensors, recording peaks. Peaks size the system, averages do not.

Step 2 — Decide what really has to stay up

Keeping everything up for six hours costs a fortune. Segment it:

  • What must never stop: network core, domain controllers, production database, carrier link. Often less than 30% of the total load.
  • What can stop cleanly: workstations, test servers, printing. A controlled shutdown beats an abrupt cut.
  • Cooling, a case of its own: a room without cooling heats up fast, and servers shut themselves down. Plan either partial autonomy on cooling, or a shutdown procedure triggered on a temperature threshold.

Step 3 — Choose between batteries, generator and solar

The three do not play the same role, and the commonest mistake is making one carry another's job.

  • The UPS covers the instant cut and the switchover time. On IT loads, aim for the autonomy needed to start the generator and shut down cleanly — not for the whole day.
  • The generator covers long outages. Its weak point is not power, it is operations: fuel on hand, automatic start tested, servicing kept up. A generator that will not start in January because it has not run since September is a textbook case.
  • Solar changes the picture on daytime outages, which are the most frequent, and cuts fuel consumption. On a site that consumes during the day, a photovoltaic installation with storage is reasoned as an investment with a measurable return, not as a green gesture.

The three mistakes that cost money

  1. Sizing on power and forgetting energy. A 10 kVA UPS says nothing about autonomy: the battery bank determines it, and it ages. A three-year-old battery does not deliver what the datasheet promised.
  2. Not monitoring energy. UPS status, load, remaining autonomy and room temperature must arrive in the same tool as everything else — for us, PRTG or ManageEngine. Otherwise you learn about the battery failure on the day you needed it.
  3. Forgetting cycles. Six outages a day means several hundred discharge cycles a year. Equipment must be chosen for that duty: it is exactly what separates an office UPS from a server-room UPS, and a starter battery from a deep-cycle one.

And supply

Correct sizing is worthless if the spare part arrives in three months. Sea freight from Asia to Pointe-Noire is counted in weeks, and a twenty-foot container commonly costs between two and four thousand dollars — which weighs on a project's schedule as much as on its budget. We factor that in upfront: equipment chosen for locally available parts, and a safety stock on critical consumables. Our installations rely on Schneider Electric and Vertiv for UPS, distribution and cooling.

What we do on a site survey

  • A week-long load survey, peaks included.
  • Autonomy calculated per segment, against the real condition of the battery bank.
  • Costed scenarios: UPS alone, UPS plus generator, adding solar — with the payback period.
  • Energy brought under monitoring, on the same footing as the servers.

Our engineers work on both sides, server room and photovoltaics — and our training covers the three levels of solar, from fundamentals to sizing and then operations, so your teams can keep the installation running after we leave.

Sources

Share this article