Virtual Datastore Tiering for VMFS Scalability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual machine file systems (VMFS) face scalability issues and resource contention in large deployments, leading to I/O failures, data corruption, and reduced operational efficiency due to the limitations of single LUN-based storage solutions and high host synchronization requirements.
Innovation Solution
Implementing a tiered configuration with virtual datastores and logical container datastores, where virtual storage objects are provisioned by the storage solution and logical containers by the hypervisor, allowing for dynamic resizing, isolation, and improved resource management, reducing resource contention and enhancing data resilience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single LUN-based VMFS datastore is used, then storage simplicity is maintained, but resource contention increases and operational efficiency decreases as the number of VMs grows
Solution Approach 1:
The patent divides the storage system into multiple independent storage policies, each backed by separate LUNs. This segmentation allows VMs to be distributed across different storage policies, reducing resource contention and improving operational efficiency while maintaining manageable complexity through the virtualization layer.
2Device complexity
If a single LUN backs a VMFS datastore, then storage configuration is simple, but multiple simultaneous storage policies become unavailable
Solution Approach 1:
The patent implements a universal storage policy framework where multiple storage policies can simultaneously back different portions of the storage solution. Each storage policy can be independently configured with different LUNs, enabling diverse storage requirements to be met within a unified virtualized storage architecture.
3Quantity of substance
If a large VMFS volume is shared across multiple hosts, then storage utilization is optimized, but I/O failures and latency increase due to synchronization requirements
Solution Approach 1:
The patent segments the large VMFS volume into multiple smaller volumes, each backed by separate LUNs and accessible to specific storage policies. This segmentation reduces the synchronization burden across hosts, minimizing I/O failures and latency while maintaining efficient storage utilization through the virtualization layer.
4Adaptability or versatility
If multiple hosts share a single VMFS volume, then resource sharing is maximized, but atomic test and set commands increase leading to scalability issues
Solution Approach 1:
The patent divides the shared storage into multiple independent storage policies with separate LUN backings. This allows different host groups to access different storage policies simultaneously without increasing atomic test and set commands, thereby maintaining cluster scalability while preserving resource sharing capabilities.
Data Source
AI summary
Enhanced datastores for virtualized environments are disclosed. A virtual datastore (e.g., a virtual volume datastore) is generated, along with a first virtual storage object (e.g., a virtual volume object) having a first storage policy. The first virtual storage object is configured into a first logical container datastore (e.g., a virtual machine file system datastore). The virtual datastore and the first logical container datastore are connected to a hypervisor in a tiered configuration, with the virtual datastore between the hypervisor and the first logical container datastore. Data is stored in the first logical container datastore. In some examples, the virtual datastore is a hybrid datastore having both a subordinate logical container datastore and a subordinate virtual storage object datastore, with the logical container datastore being provisioned by the hypervisor and the virtual storage object being provisioned by the storage solution.