Virtual Machine Allocation Using Network Load Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing datacenter allocation and re-location methods for virtual machines do not consider communication relations and network load, leading to inefficient usage of network resources and suboptimal performance.
Innovation Solution
A method that collects network load information and determines a target allocation of virtual machines based on network load profiles, using profile creator and evaluator functions to optimize VM placement across hosts, considering communication patterns and resource availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If VMs are allocated based on hardware resource demands (CPU or memory) using existing solutions like VMware DRS, then resource usage is balanced across hosts, but network load and communication efficiency are not optimized
Solution Approach 1:
The patent extends the VM allocation parameters beyond traditional hardware resources (CPU, memory) to include network load metrics. By monitoring network throughput, packet loss, and latency, the system dynamically adjusts allocation decisions based on communication patterns between VMs, thereby optimizing network resource efficiency while maintaining hardware resource balance.
Solution Approach 2:
The system continuously monitors network load information and communication patterns between VMs, using this feedback to dynamically adjust allocation decisions. The network monitoring component collects real-time data on network performance, which feeds into the allocation algorithm to optimize VM placement based on actual communication needs rather than static hardware resource considerations alone.
2Productivity
If VMs are placed equally across hosts to balance hardware resources, then individual host resource utilization is optimized, but overall network throughput and datacenter performance deteriorate
Solution Approach 1:
The patent applies local quality by allowing different allocation strategies for different hosts based on their network characteristics and the communication patterns of VMs placed on them. Rather than uniform allocation across all hosts, the system tailors placement decisions to local network conditions, communication affinities between specific VMs, and host-specific performance characteristics, thereby optimizing both host utilization and network throughput.
3Adaptability or versatility
If dynamic re-location of VMs is activated based on changing resource demands, then resource allocation adapts to workload changes, but network load and energy consumption increase due to frequent migrations
Solution Approach 1:
The system implements dynamic VM re-location that adapts to changing workload conditions while considering network load and energy consumption. The allocation algorithm continuously evaluates current network conditions, communication patterns, and resource utilization to determine optimal migration timing, balancing the need for adaptability with the cost of frequent migrations in terms of energy consumption and network disruption.
Data Source
AI summary
A method of operating a datacenter comprising a plurality of hosts coupled by a network, the hosts being configurable to run a plurality of virtual machines, the method comprising: collecting network load information indicating a load of the network; determining a target allocation of one of the virtu machines at one of the hosts based on the collected network load information; and allocating the one virtual machine at the one host based on the determined target allocation.