Makeup Water Storage Tank for Data Centers: Redundancy and Uptime Design

A makeup water storage tank for data centers is a reliability component, not a passive reservoir. Cooling towers lose water continuously to evaporation, drift, and blowdown, and if make-up fails, the cooling loop degrades and IT load trips. The data-center make-up tank must therefore be designed around uptime: dedicated capacity, redundant pumps, power-independent filling, and monitored levels. This article covers the redundancy design that separates a data-center make-up tank from a generic industrial one.

Data-center make-up demand tracks tower losses, roughly 1.5-2 percent of circulating flow per day across evaporation, drift, and blowdown. Size the tank for several days of buffer so a supply or power interruption does not reach the cooling loop, and add margin for the hottest week. The tank is insurance against downtime, so under-sizing is the real risk.

• ~1.5-2% flow/day loss.
• Several days of buffer.
• Margin for hot weeks.
• Under-sizing is the risk.

Fill must survive a pump or feeder failure, so specify N+1 fill pumps, each on an independent power path, ideally one on UPS or generator. The tank should be able to keep filling through a site power event, because a make-up tank that depends on the same feeder as the load it protects is not redundant.

• N+1 fill pumps.
• Independent power paths.
• One on UPS / generator.
• No shared single feeder.

A data-center make-up tank that connects to potable mains must not back-contaminate it, so an air gap or reduced-pressure zone per EN 1717 (or local equivalent) is mandatory, and the lining should meet NSF-61 where the water is potable-grade. Segregation protects both the supply and the site compliance.

• EN 1717 air gap / RPZ.
• NSF-61 lining if potable.
• Protect supply and compliance.
• Segregation is mandatory.

The tank should report level and fill flow to the BMS with alarms at low and high points, so operators see a failing make-up path before the cooling loop feels it. Unmonitored buffer capacity is capacity you cannot trust during an event. A simple level transmitter and alarm beats a bigger silent tank.

• Level + flow to BMS.
• Low / high alarms.
• Trust comes from monitoring.
• Simple beats silent.

We have no data-center project in this library. The clean-water tanks we cite (Guadeloupe potable GFS 2024, Namibia 67,380 m3 2022, Cilincing 21,088 m3 2024) are analogous critical-infrastructure storage, not data-center deliveries, offered as evidence of our ability to build large clean coated storage.

• No data-center project.
• Citations analogous, not claims.
• Clean storage build is real.
• Uptime design needs your spec.

Several days of buffer based on ~1.5-2 percent daily circulating-flow loss from evaporation, drift, and blowdown, plus margin for the hottest week.

Fill must survive a pump or feeder failure; N+1 pumps on independent power, ideally one on UPS or generator, keep the cooling loop supplied through a site event.

Yes, an air gap or reduced-pressure zone per EN 1717 and a potable-grade lining (NSF-61) where it connects to potable mains, to avoid back-contamination.

Level and fill flow to the BMS with low and high alarms, so a failing make-up path is seen before the cooling loop is affected.

A clean inert lining (GFS fired 820-930 C or FBE), or 316 stainless for compact loops, because the water may be potable-grade.

Not in our project library. We cite clean-water tanks at other critical sites as analogous capability, not data-center deliveries.

A makeup water storage tank for data centers is a reliability asset, so design it around uptime: dedicated multi-day buffer, N+1 power-independent fill pumps, EN 1717 backflow segregation, and BMS monitoring with alarms. The lining and volume matter, but the redundancy, the independent power, and the visibility are what keep a tower make-up failure from becoming an IT outage. Treat the tank as a monitored, redundant component and it quietly protects the cooling loop; treat it as a passive reservoir and it becomes the single point of failure you never saw coming.