As memory technology scales down, memory cells tend to become unreliable. To guarantee reliable operations of circuits and systems, semiconductor devices are tested with accelerated life tests to estimate device-level reliability and to develop a predictive reliability model for circuits and systems based on the estimated reliability of devices. However, to accurately estimate the lifetime of a system, accelerated lifetime testing at the system level is necessary. In this paper, we propose a system-level accelerated life test plan. As a case study, we investigate the reliability of SRAMs under time-dependent dielectric breakdown, one of the major contributors to aging failures in modern computer systems. Using the simulation results, we examine failure trends at various stress conditions for the system-level accelerated life test. From the resulting observations, we suggest a method that optimizes the experimental design of system-level accelerated life tests.