Micro VM Instantiation Using Offloaded Networking and Swappable Intermediaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Designing a virtualization service that can efficiently manage large, dynamically changing mixes of virtual machines with widely differing functional and performance requirements remains a non-trivial technical challenge, particularly in managing short-lived and small-resource-footprint micro virtual machines.
Innovation Solution
A virtualized computing service (VCS) optimizes resource utilization by employing virtualization offloading cards, multiplexed virtual network interfaces, and intelligent storage management to efficiently instantiate and manage large numbers of micro virtual machines, supporting both standardized and flexible-capacity VMs, with features like multiplexed VNIs, isolated virtual networks, and swappable virtualization intermediary processes to enhance responsiveness and resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional virtualization management is used to handle large numbers of virtual machines, then the system can support diverse virtual machine requirements, but the CPU, memory, and storage resource overhead increases significantly
Solution Approach 1:
The patent segments virtualization management into two distinct paths: a fast path for micro virtual machines that bypasses traditional virtualization overhead, and a standard path for conventional virtual machines. This segmentation allows the system to handle diverse VM requirements while minimizing resource consumption for the common case of short-lived micro VMs.
Solution Approach 2:
The patent introduces an intermediary mechanism (the fast path with pre-configured templates and streamlined instantiation) that mediates between the diverse requirements of virtual machines and the limited host resources. This intermediary enables efficient handling of micro VMs without requiring full traditional virtualization overhead for each instance.
2Ease of operation
If traditional virtual machine instantiation processes are used, then comprehensive configuration and management is achieved, but the responsiveness to configuration requests decreases
Solution Approach 1:
The patent applies preliminary action by pre-configuring virtual machine templates with common settings and configurations before they are needed. When a micro VM needs to be instantiated, the system can quickly apply these pre-prepared configurations rather than setting everything from scratch, thereby maintaining comprehensive management capability while dramatically improving instantiation speed.
Solution Approach 2:
The system dynamically selects between two different instantiation paths based on the virtual machine type and requirements. For micro VMs, the fast path with template-based instantiation is used for rapid deployment. For conventional VMs requiring comprehensive customization, the standard path is taken. This dynamic approach optimizes responsiveness while preserving full configuration capability when needed.
3Productivity
If separate management approaches are used for long-running and short-lived virtual machines, then each can be optimized for its specific requirements, but the overall system complexity increases
Solution Approach 1:
The patent creates a universal virtualization management system that handles both micro VMs and conventional VMs through a unified architecture. The fast path and standard path are integrated into a single management framework that automatically routes requests appropriately. This multi-functional approach allows optimization for different VM types while maintaining manageable overall system complexity through unified control.
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
An administrative agent running at a virtualization host of a network-accessible virtualized computing service determines that a first virtual machine is to be instantiated. The agent initiates at least a first configuration operation to enable connectivity for at least a portion of network traffic associated with the first virtual machine. The first configuration operation is performed at least in part using a first virtualization offloading card of the virtualization host. The agent causes a virtualization intermediary process of the virtualization host to launch one or more execution threads of the virtualization intermediary process to implement the first virtual machine. The intermediary process may be swapped to persistent storage, e.g., based on an analysis of resources of the virtualization host.