Policy Driven IO Throughput Controller for Hyper-Converged Storage

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Hyper-converged infrastructure systems face challenges in efficiently managing storage input/output (IO) throughput and network usage, particularly when multiple IO-intensive clients or applications contend for shared resources, leading to inadequate disk IO throughput and network bandwidth for critical applications.

Innovation Solution

A policy-driven IO throughput management system that adjusts IO throughput by converting client requests into network policies and controlling access to a shared pool of storage devices, using a centrally located IO throughput controller that considers global rules, performance metrics, service agreements, and time-based factors to allocate resources dynamically.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple IO-intensive clients share the same storage resources in a hyper-converged infrastructure, then resource utilization is improved, but disk IO throughput and network bandwidth for critical applications deteriorate

Engineering Contradiction:
Improveresource utilizationVSAvoiddisk IO throughput for critical applications
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements quality of service (QoS) policies that assign different priority levels to different clients based on their criticality. Critical applications receive higher priority and guaranteed minimum IO throughput, while non-critical applications receive lower priority. This local differentiation of service quality resolves the contradiction by ensuring critical applications maintain reliable performance even when multiple clients share the storage resources.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system dynamically adjusts IO throughput allocation based on real-time conditions, client priorities, and available resources. The QoS policies are not static but can be modified to respond to changing workloads and requirements. This dynamic adjustment allows the system to maintain optimal resource utilization while ensuring critical applications receive adequate throughput when needed.

Inventive Principle:
Principle #15Dynamics

2Ease of operation

If a centralized storage controller is used to manage IO throughput, then IO throughput allocation control is improved, but device complexity increases

Engineering Contradiction:
ImproveIO throughput allocation controlVSAvoidcontroller architecture
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent integrates multiple functions into a unified QoS management system that handles policy definition, monitoring, enforcement, and adjustment. The centralized controller performs multiple roles including resource allocation, performance monitoring, policy management, and dynamic adjustment, reducing the need for separate specialized components and simplifying the overall architecture while maintaining effective control.

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

Solution Approach 2:

The system automatically monitors resource usage, detects violations of QoS policies, and adjusts IO throughput allocation without requiring manual intervention. The centralized controller self-manages the complex tasks of policy enforcement and dynamic resource reallocation, reducing operational complexity despite the sophisticated control mechanisms in place.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS10379771B1Systems and methods for policy driven storage in a hyper-convergence data center
Publication Date: 2019.08.13 EMC IP HLDG CO LLC
  • US10379771B1 patent drawing
  • US10379771B1 patent drawing
  • US10379771B1 patent drawing

AI summary

Aspects of the present disclosure involve a system architecture for a policy driven disk IO throughput control for a hyper-converged storage provider. The computing architecture provides a flexible and real-time feature to the IO throughput management of a hyper-converged or converged infrastructure. In particular, through the use of centrally applied policy driven controls, the disk IO throughput allocation of different applications/clients of the converged infrastructure is gauged or otherwise controlled over the network bandwidth that link to the storage pool of the infrastructure. Through the use of the system architecture, the converged infrastructure may not utilize hard-coded disk resource allocation for each application/client in an isolated fashion, thereby allowing the IO throughput management to be flexible and agile in response to executed applications. Further, the IP throughput controlling and storage IP capacity of the converged infrastructure may be maintained separately.