Dynamic Erase Suspend Adaptation for SSD Workload Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing SSD controllers face performance latency and increased error risks due to frequent erase suspend operations, which disrupt the Quality of Service (QoS) and can lead to word line disturb in NAND devices, especially during high workload conditions.
Innovation Solution
Implementing a dynamic adaptation mechanism that computes a workload statistic based on read operations over time, allowing a maximum number of erase suspend operations during erase pulses when the workload is favorable and preventing them when unfavorable, thereby optimizing the number of erase suspends based on the current workload.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If erase suspend operations are frequently performed to prioritize read/write requests, then QoS for read/write operations is improved, but erase time loss increases and word line disturb errors occur
Solution Approach 1:
The patent implements dynamic adaptation of erase suspend operations by computing workload statistics and adjusting the number of allowed erase suspends based on current workload conditions. The controller transitions from static erase suspend behavior to dynamic control, modifying erase operation parameters in real-time based on workload intensity and type, thereby resolving the contradiction between maintaining QoS and minimizing erase time loss.
Solution Approach 2:
The patent changes the parameter of erase suspend operation frequency based on workload statistics. By computing metrics such as read/write ratios and queue depths, the system adjusts the number of erase suspends allowed during erase pulses, transforming the fixed parameter into a variable that adapts to workload conditions, thus balancing QoS requirements with erase operation efficiency.
2Ease of operation
If erase suspend operations are performed during ramp up of erase voltage, then read/write operations can be executed, but word line disturb errors increase
Solution Approach 1:
The patent applies preliminary anti-action by preventing erase suspend operations during the vulnerable ramp-up phase of erase voltage. The controller identifies critical time windows where word line disturb is most likely to occur and proactively disables erase suspends during these periods, counteracting potential errors before they can manifest, thus maintaining reliability while still allowing read/write operations during safer time windows.
Solution Approach 2:
The system performs preliminary analysis of workload statistics and erase operation timing to determine optimal moments for allowing erase suspends. By pre-computing workload metrics and predicting when erase operations are least vulnerable to disturb errors, the controller can safely permit read/write operations during these pre-identified safe windows, resolving the contradiction between operational ease and reliability.
3Reliability
If the number of erase suspend operations is increased to handle read/write requests, then QoS is improved, but erase operation completion time increases
Solution Approach 1:
The patent implements partial action by allowing a controlled, limited number of erase suspend operations based on workload statistics rather than permitting unlimited suspends. The controller computes optimal suspend thresholds that provide sufficient QoS for read/write operations while preventing excessive interruptions to erase operations, thus achieving partial satisfaction of both competing requirements.
Solution Approach 2:
The system employs feedback mechanisms by continuously monitoring workload statistics and adjusting the number of allowed erase suspends accordingly. The controller uses feedback from workload analysis to dynamically modify erase operation behavior, increasing suspends when workload is light and decreasing them when workload is heavy, thereby balancing QoS improvement with erase operation completion time.
Data Source
AI summary
Aspects relate to dynamically adapting the number of erase suspend operations on a non-volatile memory (NVM) based on the workload. In some aspects, erase suspend optimization involves computing a workload statistic based on at least read operations performed on the NVM over time, setting a maximum number of erase suspend operations allowed to be performed when the workload statistic compares favorably to a workload threshold, and preventing erase suspend operations from being performed when the workload statistic compares unfavorably to the workload threshold.