What Is Coolant Pump Redundancy?_
Liquid cooling loops for dense GPU racks use multiple pumps arranged so that loss of a single unit does not halt flow through the cold plates. The remaining pumps continue to move fluid at a level sufficient to remove heat until the failed unit is replaced. Automatic controls detect the change in pressure or flow and adjust operation without manual intervention.
Technical Details
Pumps are typically mounted in parallel within a coolant distribution unit or rack manifold. Isolation valves allow one pump to be serviced while the loop stays pressurized. Flow and pressure sensors feed status to the rack management controller, which can raise alerts or change pump speeds. The design avoids single points of failure in the mechanical circulation path.
How Leviathan Systems Works with Coolant Pump Redundancy
During commissioning of liquid-cooled racks the crew confirms that pump failover occurs and flow remains within acceptable limits before handing the rack over for GPU installation. The check is performed after manifold connections are made and before power is applied to the compute nodes.
Related service
Liquid Cooling Services →