NVMe Switch Storage Partition Allocation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing solutions for pooling storage resources in data centers introduce latency due to RAID controller translations, which becomes significant with high-speed NVM technology, as they require processing of I/O commands through RAID and SSD controllers.

Innovation Solution

The described solution involves configuring a switch to dynamically allocate storage device partitions to host nodes using separate device controllers or virtual functions, allowing direct communication between host nodes and storage devices through bus interface protocols, thereby bypassing the need for RAID controllers and reducing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If RAID controller translation is used to pool storage resources, then storage resource aggregation is achieved, but I/O latency increases

Engineering Contradiction:
Improvestorage resource aggregationVSAvoidI/O latency
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent extracts the RAID controller translation layer from the storage access path. By allowing host nodes to directly access storage devices through NVMe protocols without RAID controller intervention, it removes the source of latency while preserving storage aggregation capabilities through direct host-to-storage communication.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces a storage pool manager as a software intermediary that handles storage resource allocation and management without requiring hardware RAID controller translation. This software layer manages aggregated storage resources while allowing direct NVMe access, eliminating the latency caused by hardware RAID controllers.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of time

If direct NVMe access is implemented, then I/O latency is reduced, but storage resource pooling capability is lost

Engineering Contradiction:
ImproveI/O latencyVSAvoidstorage resource pooling
Core Design Contradiction:
Loss of timeVSAdaptability or versatility

Solution Approach 1:

The patent makes the storage system universally accessible by implementing multiple NVMe namespaces that can be shared among multiple host nodes. Each namespace can be accessed directly by authorized hosts, providing both direct access performance and resource pooling capability through the namespace sharing mechanism.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The storage pool manager acts as a software intermediary that enables resource pooling without requiring hardware RAID controllers. It manages namespace allocation, permissions, and aggregation logic, allowing direct NVMe access while maintaining storage pooling functionality through intelligent software management.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Ease of operation

If RAID abstraction layer is used, then storage management is simplified, but I/O performance deteriorates

Engineering Contradiction:
Improvestorage managementVSAvoidI/O performance
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The patent replaces the hardware-based RAID controller mechanical system with a software-based storage pool manager. This substitution eliminates the performance penalty of hardware translation while maintaining simplified storage management through software abstraction. The NVMe protocol's native capabilities provide the performance boost without sacrificing management simplicity.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The storage pool manager serves as a software intermediary that provides simplified storage management interfaces to hosts while maintaining direct NVMe access paths. It handles the complexity of resource aggregation and allocation in software, allowing hosts to access storage directly with high performance while the manager simplifies overall storage management tasks.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS11086520B2Method and apparatus for dynamically allocating storage resources to compute nodes
Publication Date: 2021.08.10 INTEL CORP
  • US11086520B2 patent drawing
  • US11086520B2 patent drawing
  • US11086520B2 patent drawing

AI summary

Provided are a method, system, computer readable storage medium, and switch for configuring a switch to assign partitions in storage devices to compute nodes. A management controller configures the switch to dynamically allocate partitions of at least one of the storage devices to the compute nodes based on a workload at the compute node.