SmartNIC Shadow Live Migration for VMs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current live migration of virtual machines in information handling systems requires CPU overhead and the presence of a hypervisor, and may be disrupted by server failures, as it relies on network communication and hypervisor management.
Innovation Solution
Implementing a smart network interface card (smartNIC) that monitors host memory for updates, copies memory pages, and uses a watchdog timer to detect failures, enabling shadow live migration with reduced hypervisor impact and continued operation even during server failures by embedding intelligence in the smartNIC.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional live migration is used, then virtual machines can be migrated between servers, but it requires CPU overhead and hypervisor management which reduces system performance
Solution Approach 1:
The patent extracts the live migration function from the CPU and hypervisor, implementing it directly in the network interface card's memory management unit. This allows migration to occur at the network layer rather than requiring CPU intervention, thereby eliminating the CPU overhead associated with traditional hypervisor-based migration while preserving the ability to migrate virtual machines between servers
Solution Approach 2:
The patent introduces the memory management unit in the network interface card as an intermediary component that handles migration operations independently. This intermediary structure enables migration to be decoupled from the host CPU and hypervisor, allowing the network interface card to manage memory pages and migration processes autonomously, thus reducing system-level overhead
2Reliability
If traditional live migration relies on network communication and hypervisor management, then migration can be coordinated centrally, but it becomes vulnerable to server failures disrupting the migration process
Solution Approach 1:
The patent implements self-service by enabling the network interface card to autonomously monitor system health through the watchdog timer and independently initiate migration operations without requiring continuous hypervisor coordination. When a failure is detected, the network interface card automatically manages the migration process, making the system resilient to server failures that would otherwise disrupt centrally coordinated migration
Solution Approach 2:
The patent applies preliminary action by having the network interface card continuously monitor system health status through the watchdog timer before failures occur. This proactive monitoring allows the system to detect potential failures and initiate migration preparations in advance, ensuring that migration can proceed smoothly even when server failures disrupt normal operation
3Ease of operation
If hypervisor manages all migration operations, then centralized control is maintained, but performance overhead increases during migration operations
Solution Approach 1:
The patent segments the migration function from the hypervisor and places it within the network interface card's memory management unit. This segmentation allows migration operations to be handled at the network layer independently, reducing the computational burden on the hypervisor and CPU while maintaining operational simplicity through automated network-based coordination
Data Source
AI summary
A smart network interface card in an information handling system monitors a local host memory associated with a computer resource for an update to a memory page in the local host memory. After the update to the memory page, the smart network interface card copies the memory page to its memory. The smart network interface card sets a watchdog timer to detect a failure in an the information handling system that hosts the computer resource and if the failure is detected, then the smart network interface card migrates the computer resource from its to another information handling system.