Unified I/O Adapter for Multi-Core Processor Bottlenecks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Multi-core processors face Input/Output (I/O) bottlenecks and processing overhead due to the discrepancy between multiple virtual and single physical resources, leading to inefficiencies in resource utilization and increased latency.

Innovation Solution

A unified hybrid adapter with an I/O interconnect that selectively switches communications between a host device, a high-latency storage device, and a low-latency storage device, caching requests in the low-latency storage and redirecting them through a network adapter for efficient processing, thereby reducing I/O bottlenecks and processing overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple virtual instances of system resources are provided to accommodate I/O requests from different operating system instances, then processor utilization is improved, but I/O bottlenecks occur because a single physical resource cannot service multiple requests simultaneously

Engineering Contradiction:
Improveprocessor utilizationVSAvoidI/O latency
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent segments the single physical storage resource into multiple virtual storage instances, each capable of being accessed independently by different operating system instances. This segmentation allows parallel access to storage resources, eliminating the bottleneck where a single physical resource could only service one request at a time, thus reducing I/O latency while maintaining high processor utilization.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a virtualization layer as an intermediary between the physical storage resource and multiple operating system instances. This virtualization layer manages and allocates storage resources dynamically, enabling multiple virtual instances to access the physical storage concurrently through virtual pathways, thereby resolving the conflict between high processor utilization and I/O latency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If virtualization is implemented to improve processor utilization, then multiple operating system instances can run concurrently, but processing overhead increases due to software routines managing virtual instances

Engineering Contradiction:
Improveprocessor utilizationVSAvoidprocessing overhead
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent merges the storage management functionality directly into the storage adapter hardware, combining virtualization capabilities with physical storage control in a single integrated unit. This integration eliminates the need for separate software routines to manage virtual instances, reducing processing overhead and energy consumption while maintaining the ability to support multiple operating system instances concurrently.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The storage adapter is designed with self-service capabilities, where the hardware itself performs virtualization and resource management tasks that would otherwise require software intervention. The storage adapter autonomously manages multiple virtual instances and allocates storage resources without requiring continuous software routine execution, thereby reducing processing overhead while maintaining high processor utilization.

Inventive Principle:
Principle #25Self-service

3Device complexity

If a single physical storage resource is used to serve multiple virtual instances, then device complexity is reduced, but I/O bottlenecks occur and latency increases

Engineering Contradiction:
Improvestorage architectureVSAvoidaccess latency
Core Design Contradiction:
Device complexityVSLoss of time

Solution Approach 1:

The patent segments the single physical storage resource into multiple virtual storage instances, creating independent access pathways for each virtual instance. This segmentation allows parallel I/O operations to occur simultaneously, reducing access latency while maintaining a relatively simple physical architecture. The virtual segmentation is achieved through firmware or hardware logic in the storage adapter rather than through complex physical multipathing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent resolves the latency issue by adding a virtualization dimension to the storage architecture. Instead of increasing physical storage resources, the solution creates a virtual layer that multiplexes access to the single physical resource, effectively adding a temporal and logical dimension to resource allocation. This allows multiple I/O operations to be serviced concurrently through virtual pathways, reducing latency without increasing physical device complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP2630579B1Unified I/O adapter
Publication Date: 2015.12.30 MARVELL WORLD TRADE LTD
  • EP2630579B1 patent drawingFigure 1
  • EP2630579B1 patent drawingFigure 2
  • EP2630579B1 patent drawingFigure 3

AI summary

Systems, methods, and other embodiments associated with a unified hybrid input/output adapter are described. According to one embodiment, an apparatus includes an Input/Output (I/O) interconnect (120) configured to connect with a host device (140) and to provide communications with the host device (140). The apparatus also includes a network adapter (110) connected to the I/O interconnect (120) and configured to communicate with a network storage (170). The apparatus includes a host adapter (120) connected to the I/O interconnect (130) and configured to communicate with a first storage device (180) and a second storage device (190). The first storage device (180) has a higher latency than the second storage device (190). The apparatus further includes a storage logic (150) configured to control the I/O interconnect (130) to cause storage access requests from the host device (140) to be cached in the second storage device (190) via the host adapter (120).