Server monitoring device for automatic error prevention in high-performance computing servers

Abstract

Server monitoring device for automatic fault prevention in a high-performance computing server, comprising: an operational data acquisition unit connected to temperature sensors, current sensors, voltage sensors, speed sensors, and flow sensors of the high-performance computing server and configured for time-related acquisition of temperature values, current values, voltage values, fan speeds, and coolant flow values; an allocation memory in which, for each computing unit, its spatial position, neighboring computing units, an assigned power supply unit, and an assigned cooling channel are stored; a limit value memory in which, for each acquired operational value, an upper limit, a lower limit, and an allowable rate of change are stored;a load evaluation unit designed to determine a temporal sequence of limit exceedances, limit falls below the limit, and exceedances of the permissible rate of change; a fault path determination unit that, based on the temporal sequence and the spatial, electrical, and cooling-related assignments stored in the allocation memory, identifies a source and the affected computing units; a control unit that generates control signals exclusively for the computing units assigned to the source and the fault path to change the cooling capacity, reduce the clock frequency, relocate computing tasks, or electrically isolate the system; and a countermeasures testing unit that, after issuing a control signal, again records operating values ​​of the computing units assigned to the fault path and, if the limit violation persists, generates a control signal for the electrical isolation of these computing units.