Hypervisor I/O Load Distribution for VM Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional systems face performance issues when instantiating multiple virtual machines due to high input/output latency caused by asynchronous operations during live migration, which impede synchronous operations in other virtual machines.
Innovation Solution
The method involves a hypervisor maintaining simultaneous connections to multiple datastores, determining the usage of these connections, and selecting the least loaded connection for new virtual machine instantiation to distribute input/output loads, thereby reducing latency and improving performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional systems instantiate multiple virtual machines using backup disk images, then business continuity is facilitated, but input/output latency increases and performance deteriorates
Solution Approach 1:
The patent segments the connection to the network-attached-storage system into multiple separate connections, each assigned to different virtual machines. Instead of multiple virtual machines sharing a single connection, each VM gets its own dedicated connection, thereby dividing the I/O load across separate communication channels and reducing latency for synchronous operations.
Solution Approach 2:
The patent transitions from a single-dimensional connection model (one connection shared by multiple VMs) to a multi-dimensional model where connections are distributed across different dimensions of the storage network. By maintaining simultaneous connections to multiple datastores and selecting connections based on usage, the system adds a new dimension of connection management that reduces I/O latency.
2Extent of automation
If asynchronous input/output operations are performed for live migration, then virtual machine restoration is enabled, but synchronous input/output operations of other virtual machines are impeded
Solution Approach 1:
The patent introduces a selection mechanism that acts as an intermediary between virtual machines and datastore connections. This intermediary monitors connection usage and intelligently assigns connections to VMs, mediating between the needs of asynchronous live migration operations and synchronous application operations by selecting the least loaded connection for each VM.
Solution Approach 2:
The patent implements dynamic connection assignment where the hypervisor continuously monitors connection usage and adapts connection allocation based on current load conditions. Instead of static connection assignment, the system dynamically selects connections for VMs based on real-time usage metrics, allowing asynchronous operations to complete while minimizing impact on synchronous operations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for instantly restoring virtual machines in high input/output load environments may include (1) identifying a hypervisor that is configurable to maintain simultaneous connections to multiple datastores of a network-attached-storage system, (2) receiving a request to activate a new virtual machine on the hypervisor, (3) determining a usage of at least one connection from the hypervisor to at least one datastore of the network- attached-storage system, and (4) selecting, based on the usage of the connection, a designated connection from the hypervisor to a designated datastore of the network-attached storage system for activating the new virtual machine on the designated datastore of the network-attached-storage system.