Direct-Attach Storage for Fault-Tolerant Virtualized Services

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Stateful applications in virtualized computing systems do not benefit from the full data protection services provided by underlying shared storage systems, leading to reduced fault tolerance due to unawareness of application replicas' placement by the storage system.

Innovation Solution

Implementing a paravirtual storage layer that utilizes direct-attach storage, allowing persistent applications to access local storage devices directly, bypassing the virtual SAN stack, and enabling controlled placement of persistent data across these devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If stateful applications use shared storage systems (SAN or vSAN) for persistent data storage, then data protection and replication services are provided by the storage system, but the applications do not benefit from the full data protection services due to the storage system's unawareness of application replicas placement

Engineering Contradiction:
Improvefault toleranceVSAvoidstorage architecture complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces Kubernetes as an intermediary layer between applications and shared storage systems. Kubernetes manages persistent volumes and their placement across hosts, coordinating with the storage system to ensure that both application-level and storage-level replication work together effectively. This intermediary enables the storage system to be aware of replica placement through Kubernetes-controlled mechanisms, resolving the fault tolerance issue while maintaining manageable complexity through abstraction.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If stateful applications perform their own replication (e.g., 3-way replication) while the underlying storage system also performs replication (e.g., 2-way replication), then multiple copies of data are created, but the fault tolerance only matches the storage system's replication level because the storage system is not aware of application replicas correlation

Engineering Contradiction:
Improvefault toleranceVSAvoiddata copies
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent implements feedback mechanisms where Kubernetes receives information about storage system replication status and uses this feedback to make intelligent decisions about persistent volume placement. The storage system provides feedback about existing replicas, and Kubernetes adjusts its placement decisions accordingly, ensuring that application replicas are placed on different hosts from storage replicas. This coordinated feedback loop eliminates redundant data copies while maintaining or improving fault tolerance.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If applications deploy on virtualized computing systems with shared storage, then flexibility in application development is provided, but efficiency is reduced due to the indirect access path through the virtual SAN stack

Engineering Contradiction:
Improveapplication deployment flexibilityVSAvoiddata access efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent segments the storage access path by introducing distinct layers: applications access persistent volumes through Kubernetes abstractions, which then access shared storage through controlled interfaces. This segmentation allows applications to maintain flexible deployment models while the Kubernetes layer optimizes the actual storage access paths. Persistent volumes can be mounted to specific pods and scheduled on specific hosts, creating directer access paths while preserving the flexibility of virtualized deployment.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS12405740B2Direct access storage for persistent services in a virtualized computing system
Publication Date: 2025.09.02 VMWARE INC
  • US12405740B2 patent drawing
  • US12405740B2 patent drawing
  • US12405740B2 patent drawing

AI summary

An example virtualized computing system includes a cluster of hosts having a virtualization layer executing thereon and configured to manage virtual machines (VMs); a local storage device in a first host of the cluster, the local storage device configured as direct-attach storage; and an orchestration control plane, integrated with the virtualization layer, the orchestration control plane including a master server managing state of the orchestration control plane, the state including objects representing the hosts and the VMs, the orchestration control plane deploying a persistent application executing on a first VM of the VMs, the persistent application storing persistent data on the direct-attach storage.