Asynchronous Checkpoint Data Transfer for Virtual Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault-tolerant computing systems face challenges in minimizing downtime during checkpointing processes, especially when dealing with multiple virtual machines, as the process is time-consuming and prone to processing slowdowns and errors due to the need for synchronized data transfer between active and standby processors.
Innovation Solution
A method for transferring checkpoint data from an active virtual machine to a standby virtual machine by breaking data into packets and sending them independently, allowing out-of-order or interleaved transmission, with a table of contents for reassembly, to minimize downtime and processing delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If checkpoint data is transferred synchronously waiting for all components, then data completeness is ensured, but checkpointing time increases and system productivity decreases
Solution Approach 1:
The patent divides checkpoint data into multiple packets, each containing data from one or more components. These packets are transmitted independently rather than waiting for all components to complete data generation, allowing asynchronous transfer while maintaining data integrity through packet tracking and reassembly mechanisms.
Solution Approach 2:
The system prepares checkpoint data packets in advance by segmenting them before transmission begins. The table of contents is generated upfront to track packet delivery status, enabling the receiving system to reassemble data efficiently once all packets arrive, rather than waiting for synchronized completion.
2Loss of time
If checkpointing interval is reduced to minimize downtime, then system availability improves, but processing errors increase due to insufficient time for complete data transfer
Solution Approach 1:
The system enables continuous transmission of checkpoint data packets without idle waiting periods between components. By transmitting packets as soon as they are prepared and tracking their delivery status, the system maintains continuous data flow, reducing total checkpointing time while ensuring all data is received before failover.
Solution Approach 2:
The table of contents structure provides feedback mechanisms to track which packets have been successfully transmitted and received. This allows the system to monitor data transfer completeness in real-time and ensure all necessary data arrives before the standby system takes over, even with reduced checkpointing intervals.
3Reliability
If multiple virtual machines are checkpointed simultaneously, then fault tolerance coverage improves, but system complexity and processing difficulty increase
Solution Approach 1:
The patent combines checkpoint data from multiple virtual machine components into a unified packet transmission system. The table of contents aggregates tracking information for all packets across all virtual machines, allowing coordinated management of multiple VM checkpointing operations through a single structured approach rather than separate independent processes.
Data Source
AI summary
In part, the invention relates to a system and method for writing checkpointing data to a computer having a standby virtual machine for each checkpointed component on a computer having an active virtual machine. In one embodiment, the checkpointing data is processed on a per virtual machine basis. This is performed in a way that allows checkpointing data packets from multiple sources to be transferred asynchronously, subsequently reassembled into a coherent checkpoint message, and applied asynchronously.