A Vulnerability Factor for ECC-protected Memory

Fault injection studies and vulnerability analyses have been used to estimate the reliability of data structures in memory. We survey these metrics and look at their adequacy to describe the data stored in ECC-protected memory. We also introduce FEA, a new metric improving on the memory derating factor by ignoring a class of false errors. We measure all metrics using simulations and compare them to the outcomes of injecting errors in real runs. This in-depth study reveals that FEA provides more accurate results than any state-of-the-art vulnerability metric. Furthermore, FEA gives an upper bound on the failure probability due to an error in memory, making this metric a tool of choice to quantify memory vulnerability. Finally, we show that ignoring these false errors reduces the failure rate on average by 12.75% and up to over 45%.

I. Introduction

Faults in DRAM cells are becoming ever more prevalent with new process technologies, causing the protection using Error Correcting Codes (ECC) to increase from strong ChipKill protection [7] to include a second layer of ECC on-chip [13]. This trend includes even mobile devices [21] and efforts to gain in energy efficiency could further multiply the soft error rates [2]. To apply more granular protection techniques than uniformly increasing ECC strength, such as dynamically adaptable ECC [27], it is necessary to quantify the risk associated with any data stored in memory.