John FragallaHPCatDell home button
Power and cooling is a critical aspect when architecting a HPC system. In HPC, there is an increase in higher memory capacity per processor core, the increase use of GPUs, and system vendors are designing higher system densities, which are increasing the number of components per rack, resulting in higher power consumption per rack. A component in a HPC system can be compute system, network switches, or storage arrays, for example.. The power consumption per rack is commonly exceeding 20 kilowatts (KW), and it some cases, 30 KW per rack, depending on the system configuration. Most data centers today are limited not just by power, but also by cooling. Today, some data centers can only power or cool less than 10KW per rack, while other data centers can power or cool up to 15KW per rack, and in some cases, can power or cool more than 20KW per rack.

There are many aspects to consider in architecting a HPC system to meet data center limitations . One must know the limiting power or cooling requirements per rack in the data center where the HPC system will be installed. Knowing the data center rack requirement will determine how many components can be installed in a single rack based on peak power consumption of those components. One method in measuring component power consumption of a system is running LINPACK. After determining the number of components that can be installed in a rack, a few other important aspects of power and cooling to consider is Amps, Cubic Feet per Minute (CFMs), Tons of Cooling, and British Thermal units per hour (BTUs/hr).

HPC system design is not just only about number of processors, memory per processor cores, Interconnects, or storage capacity and throughput, it’s also about determining the power and cooling aspect of the system to meet the data center requirements, which will result in a more power efficient and reliable system.

Read HPC Power and Cooling: Power Consumption – Part 2

-- John Fragalla