SSD Wear Leveling via Host-Controlled Data Relocation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Direct placement SSDs face challenges in managing wear leveling and reducing uncorrectable bit errors (UBER) without impacting latency and performance, as existing methods either require complex host integration or tie up SSD resources with internal data movements.

Innovation Solution

The SSD manages media policies internally and communicates abstracted policies to the host, allowing the host to control background data movements opportunistically, ensuring wear leveling and UBER reduction without sharing detailed policies, thus maintaining performance and reducing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the SSD performs internal data movements for wear leveling, then wear leveling is achieved, but SSD resources are tied up and latency increases

Engineering Contradiction:
Improvewear levelingVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts the wear leveling data movement operations from the SSD and transfers them to the host system. The SSD controller identifies blocks requiring wear leveling and sends notifications to the host, which then performs the actual data relocation operations. This separation allows the SSD to maintain low-latency operations while the host handles background wear leveling tasks during its available cycles.

Inventive Principle:
Principle #2Taking out (Extraction)

2Ease of operation

If the SSD shares detailed media policies with the host, then the host can manage wear leveling, but host integration complexity increases

Engineering Contradiction:
Improvehost controlVSAvoidhost integration
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent creates a simplified copy of the media policy information that the host needs to perform wear leveling. Instead of requiring the host to integrate with complex SSD internal policies and mechanisms, the SSD controller translates and communicates only the essential block identification and relocation requirements to the host. This abstraction layer maintains ease of operation while minimizing host integration complexity.

Inventive Principle:
Principle #26Copying

3Productivity

If the host performs background data movements, then wear leveling is achieved without SSD resource tie-up, but host processing cycles are consumed

Engineering Contradiction:
ImproveSSD performanceVSAvoidhost processing
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent implements a self-service mechanism where the host system utilizes its own available processing cycles and resources to perform wear leveling operations on the SSD. The SSD controller identifies when wear leveling is needed and notifies the host, which then autonomously performs the data relocation using its spare processing capacity. This approach improves SSD productivity without requiring continuous SSD resource allocation, as the host services its own storage maintenance needs during idle periods.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS10254979B1Relocating or aborting a block of data by a host, based on media policies managed by a storage device
Publication Date: 2019.04.09 SK HYNIX NAND PRODUCT SOLUTIONS CORP
  • US10254979B1 patent drawing
  • US10254979B1 patent drawing
  • US10254979B1 patent drawing

AI summary

A storage device transmits a message to a host, where the message indicates that at an opportune time the host should relocate or abort a block of data that is stored in the storage device. The storage receives a command from the host, to relocate or abort the block of data stored in the storage device.