An Early-Life NAND Flash Endurance Prediction System

NAND flash memory – ubiquitous in today’s world of smart phones, SSDs (solid state drives), and cloud storage has a number well-known reliability problems. data contains bit errors, which require the use error correcting codes (ECCs). The raw rate (RBER) increases with program-erase (P-E) cycling, P-E cycles device can withstand before RBER exceeds ECC capability is called its endurance. operates on stored sector NAND, there large variation endurance sectors within across devices, resulting excessively conservative specifications. This research shows, for first time, that sector’s true be predicted remarkable accuracy, using combination location device, measurements taken at very beginning life. Real-world gathered millions custom-built test platform. Optimised machine learning classification models are built from to predict if will pass or fail fixed threshold, after target cycling level been reached. A novel technique demonstrated uses different thresholds model training testing, allows tuned so they never misclassify samples would fail. eliminates failures loss, allowing simpler, less expensive schemes used modern devices. It also enables significant extensions achieved.