VM Migration via Network Affinity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed computing resource clusters, peripheral devices are often not available to virtual machines or containers running on other host systems, even within the same cluster, leading to inefficient network communications and resource utilization.
Innovation Solution
A network-efficient isolation environment redistribution method is implemented, where network communications among isolation environments like VMs and containers are surveyed to identify affinities, allowing a distributed resource scheduler to migrate these environments and reduce network traffic by collocating containers and VMs on the same host machine based on network-level context information and application-affinity policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines and containers are distributed across multiple host machines in a cluster, then resource utilization is improved, but network communication overhead increases and peripheral device availability decreases
Solution Approach 1:
The system implements network context collection and analysis to monitor communication patterns between isolation environments. The distributed resource scheduler uses this feedback information about network traffic patterns to make intelligent redistribution decisions, moving containers with high communication affinity to the same host machine, thereby reducing network overhead while maintaining efficient resource utilization.
Solution Approach 2:
The system dynamically redistributes containers and virtual machines based on changing network communication patterns and resource utilization metrics. The distributed resource scheduler continuously evaluates network context information and adjusts the placement of isolation environments in real-time, transforming the static allocation into a dynamic optimization process that adapts to workload changes.
2Productivity
If isolation environments are redistributed to reduce network traffic, then network communication efficiency is improved, but system complexity increases
Solution Approach 1:
The system segments the resource management functionality into a distributed resource scheduler operating on multiple host machines within the cluster. Each scheduler instance independently manages local containers and makes redistribution decisions based on network context information, dividing the complex optimization problem into manageable local decisions rather than requiring centralized control of all systems.
Solution Approach 2:
The patent introduces network context information as an intermediary that mediates between the need for resource distribution and the desire to reduce network traffic. The distributed resource scheduler uses this intermediate data structure to make informed decisions about container placement, avoiding direct complex interactions between all components while achieving optimized resource allocation.
Data Source
AI summary
Network-efficient isolation environment redistribution is described. In one example, network communications are surveyed among isolation environments, such as virtual machines (VMs) and containers, hosted on a cluster. An affinity for network communications between the isolation environments can be identified based on the survey. Pairs or groups of the isolation environments can be examined to identify ones which have an affinity for network communications between them but are also hosted on different host machines in the cluster. The identification of the affinity for network communications provides network-level context for migration decisions by a distributed resource scheduler. Certain VMs and/or containers can then be migrated by the distributed resource scheduler to reduce the network communications in the cluster based on the network-level context information.