Design of Server Fault Diagnosis and Prediction Based on BMC

The server has become the central part of the information society, which means that the stable operation of the server is essential for people. Hence, the server fault diagnosis and prediction based on the baseboard management controller (BMC) is designed in this study to help administrators manage the server. The design scheme is divided into two phases according to the fault handling of the central processing unit (CPU). In the system startup phase, the basic input output system (BIOS) collects the errors in the system startup and sends them to BMC. In the running phase, the errors from the CPU, Memory, and PCIE are processed by the Reliability, Availability, and Serviceability (RAS) module but the errors from the NVME and PSU are handled by the registers. Hence, BMC obtains and handles the data by different methods. Then, the frequency reduction and DTF algorithms are adopted to maintain normal system status when BMC detects the alarm events. In summary, the novel design of fault diagnosis and prediction in this study contains a complete set of solutions for fault data collection, fault diagnosis, and fault prediction. Based on this, the design introduced by this paper can improve server management efficiency and reduce the probability of failure occurrence.