VM Cache Migration with Conditional Read Redirection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing virtual machine (VM) migration techniques face challenges in maintaining seamless operation and reducing latency during cache migration, as they often require clearing cache data, leading to high latency and slow cache warming at the destination host.

Innovation Solution

A system and method for migrating VM cache by sending cache data from one host to another upon notification of migration, allowing read and write operations to be processed without waiting for complete data transfer, with conditional redirection to the original cache and prioritizing most recently accessed data for efficient transfer and reduced bandwidth overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If cache data is cleared during VM migration, then the destination cache can be populated with transferred data, but data access latency increases significantly during the cache warming process

Engineering Contradiction:
Improvecache data availabilityVSAvoidcache warming time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by maintaining and transferring cache data from the source host to the destination host before the VM migration is complete. The cache data is transferred in advance and made available at the destination, so that when the VM arrives, the cache is already warm and ready to serve requests without latency penalties.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The cache data transfer operates continuously throughout the migration process rather than being performed as a discrete post-migration step. Data transfer and VM migration proceed in parallel, ensuring that the cache warming process does not interrupt or delay the migration, and that cache data is continuously made available at the destination.

Inventive Principle:
Principle #20Continuity of useful action

2Reliability

If all cache data is transferred from source to destination host, then the destination cache is fully populated, but network bandwidth consumption increases

Engineering Contradiction:
Improvecache data completenessVSAvoidnetwork bandwidth
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The system applies local quality by transferring only the specific cache data that is relevant to the migrating VM, rather than transferring all cache data from the source host. The cache data transfer is tailored to the local needs of the destination VM, sending only the portion of cache data that will be actually used, thus reducing unnecessary network bandwidth consumption.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system performs partial action by transferring a subset of cache data that is sufficient to warm the destination cache for the VM's expected workload, rather than transferring the complete cache dataset. This partial transfer provides enough data to achieve low latency for the VM's actual access patterns without the overhead of transferring excess data.

Inventive Principle:
Principle #16Partial or excessive action

3Speed

If read operations are redirected to the original cache during migration, then data can be served immediately, but the complexity of managing dual cache access increases

Engineering Contradiction:
Improvedata access speedVSAvoidcache access management
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The system introduces an intermediary mechanism (the cache controller or VMM) that manages the redirection of read operations between source and destination caches. This intermediary automatically routes read requests to the appropriate cache based on the migration state, shielding the VM from the complexity of dual-cache management while maintaining fast data access.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system uses feedback by monitoring the migration state and cache data transfer progress to dynamically adjust read operation routing. As cache data is transferred and becomes available at the destination, the system receives feedback about this state change and automatically updates the routing of read operations to prioritize the destination cache, maintaining optimal access speed while simplifying management.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3350713B1Distributed cache live migration
Publication Date: 2022.05.04 HUAWEI TECH CO LTD
  • EP3350713B1 patent drawingFigure 1
  • EP3350713B1 patent drawingFigure 2
  • EP3350713B1 patent drawingFigure 3

AI summary

A system, comprising: a first host comprising a first cache and associated with a virtual machine, VM; and a second host comprising a second cache; wherein the first host is adapted to send cache data of the first cache to the second host in response to a notification, said cache data associated with the VM and said notification indicating that the VM is to be migrated from the first host to the second host, and wherein the first host is adapted to send write operations associated with the VM to the second host in response to receiving the notification; and wherein the second host is adapted to apply, in response to receiving the notification, read operations associated with cache data of the VM to the first cache if said cache data is not present in the second cache, and to apply write operations associated with cache data of the VM to the second cache.