Flash Memory Internal Bus Segmentation for Garbage Collection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current flash memory systems face performance bottlenecks due to garbage collection and scrub operations that lock up mainline bus bandwidth, reducing overall efficiency and data integrity.

Innovation Solution

Implementing a method and apparatus that utilize an internal bus for garbage collection and scrub operations within a flash memory chip, with a function engine performing these tasks and an on-flash chip memory buffer, allowing for interleaved operations with mainline reads and writes, and a switching infrastructure to manage bus usage effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If garbage collection and scrub operations use the mainline bus for data movement, then these operations can be performed, but I/O bandwidth is locked up and overall performance reduces significantly

Engineering Contradiction:
Improvedata integrityVSAvoidI/O bandwidth
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the bus system into two distinct channels: the mainline bus for host I/O operations and an internal bus for garbage collection and scrub operations. This segmentation allows GC and scrub to execute independently on the internal bus without competing for bandwidth with mainline reads and writes, thereby preserving I/O performance while maintaining data integrity through background maintenance operations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an internal bus as an intermediary channel between the flash memory controller and memory blocks. This intermediary allows GC and scrub operations to transfer data without occupying the mainline bus, effectively mediating between the need for maintenance operations and the need for high I/O throughput.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If garbage collection operations are performed frequently to maintain free blocks for writes, then write performance is maintained, but mainline bus bandwidth is consumed reducing overall efficiency

Engineering Contradiction:
Improvewrite performanceVSAvoidbus bandwidth consumption
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

By segmenting the bus into mainline and internal channels, the patent enables GC operations to consume no mainline bus bandwidth. Write performance is maintained through background GC execution on the internal bus, while mainline bandwidth remains fully available for host I/O operations, eliminating the trade-off between write performance and bandwidth consumption.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs garbage collection as a preliminary background operation that proactively frees up blocks before they are needed for writes. By executing GC in advance on the internal bus without blocking mainline operations, the system ensures free blocks are available when needed while avoiding the performance penalty of synchronous GC execution.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If scrub operations are performed to catch and correct errors preemptively, then data integrity is ensured, but read bandwidth is locked up by the flash controller

Engineering Contradiction:
Improvedata integrityVSAvoidread bandwidth
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the data path for scrub operations to use the internal bus rather than the mainline bus. This allows scrub to read from and write to memory blocks without occupying read bandwidth on the mainline bus, ensuring data integrity through error detection and correction while leaving read bandwidth fully available for host read operations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The internal bus serves as an intermediary channel that enables scrub operations to access memory blocks without interfering with mainline read operations. This intermediary path allows the flash controller to perform preventive error checking and correction while maintaining full read bandwidth availability for host I/O.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS9400603B2Implementing enhanced performance flash memory devices
Publication Date: 2016.07.26 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US9400603B2 patent drawing
  • US9400603B2 patent drawing
  • US9400603B2 patent drawing

AI summary

A method and apparatus for implementing enhanced performance in a flash memory system in a computer system. A flash memory chip includes a function engine performing garbage collection and scrub operations using an internal bus for data movement. The system includes an on-flash chip memory buffer buffering garbage collection and scrub requests. Garbage collection and scrub operations are interleaved with mainline reads and writes.