Intel Corporation
HARDWARE RELIABILITY DIAGNOSTICS AND FAILURE DETECTION VIA PARALLEL SOFTWARE COMPUTATION AND COMPARE

Abstract:

Methods, apparatus, and software for hardware reliability diagnostics and failure detection via parallel software computation and compare. Parallel testing is performed on hardware resources such as processor cores, accelerators, and Other Processing Units (XPUs) using test algorithms such as encryption/decryption. The results of the testing (the algorithm outputs) are compared to detect errant hardware. Comparison may be across cores (via execution of software-based algorithms), across accelerators/XPUs (via algorithms implement in hardware) or between cores and accelerators/XPUs. Techniques are disclosed to enable all cores to be tested while a platform is performing a workload, such as in a data center environment, wherein unused cores are used for testing, with workloads being migrated between cores between tests.