Data Retention in MLC NAND Flash Memory: Characterization, Optimization, and Recovery Yu Cai, Yixin Luo, Erich F. Haratsch*, Ken Mai, Onur Mutlu Carnegie Mellon University, *LSI Corporation 1
You Probably Know • Many use cases: + High performance, low energy consumption 2
NAND Flash Memory Challenges – Requires erase before program (write) – High raw bit error rate Controller Raw Flash Flash CPU Memory Chips ECC Controller 3
Limited Flash Memory Lifetime Goal: Extend flash memory lifetime Raw bit error rate (RBER) at low cost P/E Cycle Lifetime ECC-correctable RBER ~2000 ~3000 Program/Erase (P/E) Cycles (or Writes Per Cell) 4
Retention Loss Charge leakage over time 0 0 1 Retention Flash cell error One dominant source of flash memory errors [DATE ‘12, ICCD ‘12] 5
Before I show you how we extend flash lifetime … NAND Flash 101 6
Threshold Voltage (V th ) Flash cell Flash cell 0 1 Normalized V th 7
Threshold Voltage (V th ) Distribution Probability Density Function (PDF) 0 1 Normalized V th 8
Read Reference Voltage (V ref ) PDF V ref 0 1 Normalized V th 9
Multi-Level Cell (MLC) ER-P1 V ref P1-P2 V ref P2-P3 V ref PDF Erased P1 P2 P3 (11) (10) (00) (01) Normalized V th 10
Threshold Voltage Reduces Over Time Before retention loss: After some retention loss: PDF P1 P2 P3 (10) (00) (01) Normalized V th 11
Fixed Read Reference Voltage Becomes Suboptimal Before retention loss: After some retention loss: P1-P2 V ref P2-P3 V ref PDF P1 P2 P3 (10) (00) (01) Normalized V th Normalized V th Raw bit errors 12
Optimal Read Reference Voltage (OPT) After some retention loss: P1-P2 OPT P2-P3 OPT P1-P2 V ref P2-P3 V ref PDF P1 P2 P3 (10) (00) (01) Normalized V th Minimal raw bit errors 13
Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage 14
Retention Failure After some retention loss: After significant retention loss: P1-P2 V ref P2-P3 V ref PDF P1 P2 P3 (10) (00) (01) Normalized V th Uncorrectable errors Correctable errors 15
Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage Goal 2: Design an offline mechanism to recover data after detecting uncorrectable errors 16
To understand the effects of retention loss: - Characterize retention loss using real chips 17
To understand the effects of retention loss: - Characterize retention loss using real chips Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage Goal 2: Design an offline mechanism to recover data after detecting uncorrectable errors 18
Characterization Methodology FPGA-based flash memory testing platform [Cai+,FCCM ‘11] 19
Characterization Methodology • FPGA-based flash memory testing platform • Real 20- to 24-nm MLC NAND flash chips • 0- to 40-day worth of retention loss • Room temperature (20⁰C) • 0 to 50k P/E Cycles 20
Characterize the effects of retention loss 1. Threshold Voltage Distribution 2. Optimal Read Reference Voltage 3. RBER and P/E Cycle Lifetime 21
1. Threshold Voltage (V th ) Distribution PDF P1 P2 P3 Normalized V th 22
1. Threshold Voltage (V th ) Distribution 0-day 0-day 40-day 40-day P1 P2 P3 Finding: Cell’s threshold voltage decreases over time 23
2. Optimal Read Reference Voltage (OPT) 40-day 0-day 40-day 0-day OPT OPT OPT OPT P1 P2 P3 Finding: OPT decreases over time 24
3. RBER and P/E Cycle Lifetime RBER P/E Cycles 25
3. RBER and P/E Cycle Lifetime V ref closer to Reading data with 7-day worth of retention loss. actual OPT Nominal Extended Lifetime Lifetime Actual OPT ECC-correctable RBER Finding: Using actual OPT achieves the longest lifetime 26
Characterization Summary Due to re retention lo loss ‐ Cell’s threshold voltage (V th ) decreases over time ‐ Optim imal read reference volt ltage (OPT) decreases over time Using the actual OPT T for reading ‐ Achieves the longest lif lifetime 27
To understand the effects of retention loss: - Characterize retention loss using real chips Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage Goal 2: Design an offline mechanism to recover data after detecting uncorrectable errors 28
Naïve Solution: Sweeping V ref Key idea: Read the data multiple times with different read reference voltages until the raw bit errors are correctable by ECC Finds the optimal read reference voltage Requires many read-retries higher read latency 29
Comparison of Flash Read Techniques Flash Read Lifetime Performance Techniques (P/E Cycle) (Read Latency) Fixed V ref Sweeping V ref Our Goal 30
Observations 1. The optimal read reference voltage gradually decreases over time Key idea: Record the old OPT as a prediction (V pred ) of the actual OPT Benefit: Close to actual OPT Fewer read retries 2. The amount of retention loss is similar across pages within a flash block Key idea: Record only one V pred for each block Benefit: Small storage overhead (768KB out of 512GB) 31
Retention Optimized Reading (ROR) Components: 1. Online pre-optimization algorithm ‐ Periodically records a V pred for each block 2. Improved read-retry technique ‐ Utilizes the recorded V pred to minimize read-retry count 32
1. Online Pre-Optimization Algorithm • Triggered periodically (e.g., per day) • Find and record an OPT as per-block V pred • Performed in background • Small storage overhead New Old PDF V pred V pred Normalized V th 33
2. Improved Read-Retry Technique • Performed as normal read • V pred already close to actual OPT • Decrease V ref if V pred fails, and retry PDF OPT V pred Normalized V th Very close 34
Retention Optimized Reading: Summary Flash Read Lifetime Performance Techniques (P/E Cycle) (Read Latency) Fixed V ref 64% ↑ Sweeping V ref 64% ↑ _____ Nom. Life: 2.4% ↓ ROR Ext. Life: 70.4% ↓ 35
To understand the effects of retention loss: - Characterize retention loss using real chips Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage Goal 2: Design an offline mechanism to recover data after detecting uncorrectable errors 36
Retention Failure After significant retention loss: After some retention loss: P1-P2 V ref P2-P3 V ref PDF P1 P2 P3 (10) (00) (01) Normalized V th Uncorrectable errors Correctable errors 37
Leakage Speed Variation PDF S S low-leaking cell F ast-leaking cell F Normalized V th 38
Initially, Right After Programming PDF P2 P3 S S F F F F S S Normalized V th 39
After Some Retention Loss Fast-leaking cells have lower V th PDF Slow-leaking cells have higher V th P2 P3 S S F F F F F F F F S S Normalized V th 40
Eventually: Retention Failure PDF OPT P2 P3 S S F F F F S S Normalized V th 41
Retention Failure Recovery (RFR) Key idea: Guess original state of the cell from its leakage speed property Three steps 1. Identify risky cells 2. Identify fast-/slow-leaking cells 3. Guess original states 42
1. Identify Risky Cells OPT –σ OPT+ σ + S = P2 Risky PDF OPT cells + F = P3 Key Formula S F F S Normalized V th 43
2. Identifying Fast- vs. Slow-Leaking Cells OPT –σ OPT+ σ + S = P2 Risky PDF OPT cells + F = P3 Key Formula ? ? ? ? ? ? Normalized V th 44
2. Identifying Fast- vs. Slow-Leaking Cells OPT –σ OPT+ σ + S = P2 Risky PDF OPT cells + F = P3 Key Formula ? S ? ? F ? ? F ? S Normalized V th 45
3. Guess Original States + S = P2 Risky PDF cells + F = P3 Key Formula S F F S Normalized V th 46
RFR Evaluation • Expect to eliminate Program with 50% of raw bit errors random data • ECC can correct remaining errors 28 days Detect failure, backup data 12 addt’l . days Recover data 47
To understand the effects of retention loss: - Characterize retention loss using real chips Goal 1: Design a low-cost mechanism that dynamically finds the optimal read reference voltage Goal 2: Design an offline mechanism to recover data after detecting uncorrectable errors 48
Conclusion Problem: Retention loss reduces flash lifetime Overall Goal: Extend flash lifetime at low cost Flash Characterization: Developed an understanding of the effects of retention loss in real chips Retention Optimized Reading: A low-cost mechanism that dynamically finds the optimal read reference voltage ‐ 64% lifetime ↑ , 70.4% read latency ↓ Retention Failure Recovery: An offline mechanism that recovers data after detecting uncorrectable errors ‐ Raw bit error rate 50% ↓ , reduces data loss 49
Data Retention in MLC NAND Flash Memory: Characterization, Optimization, and Recovery Yu Cai, Yixin Luo, Erich F. Haratsch*, Ken Mai, Onur Mutlu Carnegie Mellon University, *LSI Corporation 50
Backup Slides 51
RFR Motivation Data loss can happen in many ways 1. High P/E cycle High temperature accelerates retention 2. loss 3. High retention age (lost power for a long time) 52
Recommend
More recommend
