OVHcloud has presented a new cooling architecture for your data centers which allows the company reduce electricity consumption for cooling by 50%, and water consumption by up to 30%. The company’s data centers that will incorporate it, known as Smart Datacenters, will combine a generation of racks with a new design also with AI functions.
In development for two years, this will be the fifth generation of the OVHcloud server rack. It allows servers to be arranged in clusters, connected in series for cooling. Additionally, the servers in each cluster remain organized in parallel, making them easier to maintain. This new design optimizes the layout of components, including bulky ones like the CDU inrack. In addition, it contributes to reducing the rack’s electrical consumption.
The rack also works with a pull-type hydraulic configuration, so that each of the servers that comprise it has the appropriate water flow and pressure for its cooling needs. Its hardware components are cooled through direct-to-chip water blocks designed by OVHcloud. These blocks dissipate heat through a closed water loop that extends to a single cooling loop throughout the data center.
The cooling module is now almost 50% more compact and is outside the rack. Capable of cooling several rows of racks, it has more than 30 sensors that monitor the water parameters of the racks: pressure, speed and temperature, to adjust cooling in real time. Sensors allow the rack to take into account its immediate surroundings and data center temperatures. Using this information, the intelligent cooling module can automatically adapt to the workload of the servers. This optimization prolongs the useful life of the equipment and helps optimize the energy consumption of the infrastructure.
The Smart Dry Cooler, located outside, is the last cooling component of the closed water circuit, which occupies half the space and has half the fans that the previous generation of equipment needed. This helps reduce cooling energy consumption by up to 50%, while reducing the ambient noise level.
By constantly analyzing their environment and the behavior of their key components, the new OVHcloud Smart Racks self-regulate and collect operational data on a daily basis. The data from the racks, refrigeration modules and Dry Coolers is entered into a data lake so that complex algorithms determine predictive behaviors that contribute to optimizing maintenance and increasing the durability of the equipment.
The system can also connect to a local weather station to enrich data lakes. The precision of the models allows the exact dosage of the volume of water necessary to inject into the cooling system with a new design, which reduces water consumption by 30%, to allow heat exchange with the outside. It does not use a recirculation circuit, so it does not have hydraulic units, such as pumps, tanks and level sensors; and reduces the complexity of maintenance infrastructure.
The predictive AI that accompanies OVHcloud Smart Racks can anticipate and react to operational dimensions at the infrastructure level, including pump speed and water flow, fan speed or valve opening to optimize configuration.
Rack performance can also be adjusted based on external constraints such as noise limits, water availability or energy cost. In this way, the algorithm can choose to consume more electricity to promote water conservation or adjust sound levels to adapt to urban environments.
The new generation of OVHcloud Smart Racks is currently being deployed in its Roubaix data centerin a room with about 60 racks and 2,000 servers, and is accompanied by its new cooling system. It is planned that the entire system will be implemented in the group’s data centers.
