Lazy Disk I/O for Virtual Machine Live Migration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing virtual machine live migration techniques face inefficiencies due to redundant network transmissions of memory pages that have been laundered to a virtual disk, leading to suboptimal disk I/O performance during the migration process.

Innovation Solution

Implementing lazy disk I/O by notifying the destination hypervisor of memory page addresses and locations on the virtual disk, allowing the hypervisor to store mappings without initially reading the contents, and intercepting page faults to retrieve data only when needed, thereby reducing unnecessary disk accesses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If memory pages are transmitted over the network during virtual machine live migration, then the virtual machine can be migrated to a destination host, but redundant network transmissions occur for pages that have been laundered to a virtual disk

Engineering Contradiction:
Improvemigration speedVSAvoidnetwork transmission overhead
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The system performs preliminary identification of memory pages that have been laundered to the virtual disk before the migration process completes. By pre-marking these pages in the page table with a source bit indicator, the system avoids redundant network transmissions of the same data, thereby improving migration efficiency and reducing network overhead.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The origin hypervisor provides feedback to the destination hypervisor about which memory pages have been laundered to the virtual disk. This feedback mechanism allows the destination hypervisor to make informed decisions about page retrieval, avoiding redundant network transmissions by utilizing the virtual disk storage for pages that don't need immediate network transfer.

Inventive Principle:
Principle #23Feedback

2Reliability

If the hypervisor reads all memory page contents from the virtual disk during migration, then complete data is available for transmission, but disk I/O performance deteriorates due to excessive accesses

Engineering Contradiction:
Improvedata completenessVSAvoiddisk I/O performance
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

Instead of reading all memory page contents from the virtual disk during migration, the system performs partial action by only reading pages that are actually needed for network transmission. Pages marked with the source bit are identified as already laundered to disk and are excluded from network transmission, thereby reducing unnecessary disk I/O operations while maintaining data completeness for pages that require transmission.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system extracts and identifies the subset of memory pages that have been laundered to the virtual disk by setting source bits in the page table. This extraction allows the migration process to separate pages that need network transmission from those stored on the virtual disk, eliminating redundant disk reads and improving overall disk I/O performance.

Inventive Principle:
Principle #2Taking out (Extraction)

3Productivity

If the hypervisor implements lazy disk I/O by storing mappings without initial reading, then unnecessary disk accesses are reduced, but page fault interception and data retrieval complexity increases

Engineering Contradiction:
Improvedisk I/O efficiencyVSAvoidhypervisor complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The hypervisor performs preliminary action by storing mappings of memory pages to their locations on the virtual disk and setting source bits in the page table, without immediately reading the page contents. This preliminary setup enables lazy loading later, where pages are retrieved from the virtual disk only when accessed, reducing unnecessary disk I/O operations while managing complexity through structured page table management.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The page table serves as an intermediary structure that tracks the location of memory pages, whether in host memory or on the virtual disk. By using the source bit flag in the page table entries, the system creates a simple mechanism to distinguish between pages that need network transmission and those stored on disk, managing the complexity of lazy disk I/O through this intermediary data structure.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS9495191B2Using virtual disk in virtual machine live migration
Publication Date: 2016.11.15 RED HAT ISRAEL
  • US9495191B2 patent drawing
  • US9495191B2 patent drawing
  • US9495191B2 patent drawing

AI summary

Systems and methods for implementing lazy disk I/O in virtual machine live migration. An example method may comprise: determining, by a first computer system, that contents of a memory page mapped into an address space of a virtual machine undergoing live migration from a second computer system to the first computer system is stored on a network-accessible virtual disk; storing, in a memory data structure, a mapping of an address of the memory page to an identifier of a location of the memory page on the virtual disk; and responsive to detecting an access to the memory page by the virtual machine, storing in a memory mapped into the address space of the virtual machine the memory page retrieved from the location of the memory page on the virtual disk.