Distributed SSD Parity Offloading for Server Bandwidth Relief

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Servers face bandwidth and performance challenges in performing parity calculations due to the heavy cost of DRAM bandwidth and CPU usage, especially when connected to multiple SSDs.

Innovation Solution

Offload parity calculations to a plurality of solid-state drives (SSDs) equipped with processors that can access and determine parity information for data stored across a host address space, allowing parallel processing and reducing the server's bandwidth burden.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the server performs parity calculations using DRAM bandwidth and CPU, then data redundancy and error detection are achieved, but the server experiences heavy bandwidth consumption and performance degradation

Engineering Contradiction:

Improvedata redundancyVSAvoidserver performance

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent extracts the parity calculation function from the server system and relocates it to the SSD. The SSD controller now performs XOR operations independently, removing the computational burden from the server's CPU and DRAM bandwidth, while still achieving the required data redundancy for error detection and recovery

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The SSD controller acts as an intermediary between the server and the storage system. It receives data from the server, performs parity calculations locally using its own processing resources, and manages the redundancy information storage, thereby mediating the computational load without requiring server involvement in the actual parity operations

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If the server is connected to multiple SSDs (16-24 SSDs), then storage capacity and redundancy are increased, but the server's bandwidth becomes insufficient to handle parity calculations

Engineering Contradiction:

Improvestorage capacityVSAvoidbandwidth consumption

Core Design Contradiction:

Quantity of substanceVSUse of energy by moving object

Solution Approach 1:

Each SSD is equipped with its own processor that can independently perform parity calculations. The SSDs serve themselves by handling their own parity computation needs without requiring external server resources, enabling the system to scale to 16-24 SSDs without proportionally increasing server bandwidth requirements

Inventive Principle:

Principle #25Self-service

3Reliability

If parity calculations are performed in series on the server, then correctness is ensured, but processing time and serialization overhead increase

Engineering Contradiction:

Improvecalculation correctnessVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the parity calculation workload by assigning different XOR operations to different SSDs. Each SSD processes a portion of the parity calculations independently in parallel, maintaining calculation correctness through deterministic XOR operations while dramatically reducing total processing time by eliminating server-side serialization bottlenecks

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12436842B2Systems and methods of distributed parity calculation offloading

Publication Date: 2025.10.07 KIOXIA CORP

AI summary

A system and a method for offloading distributed parity calculations from a server including a plurality of solid-state drives (SSDs) in communication with a host via a host interface. Each SSD includes a processor. The processor of each SSD may be configured to access a host address space via the host interface, determine parity information for data stored across a plurality of buffers in the host address space, and output the parity information to the host interface.