Hypervisor Memory Reclamation in Disaggregated Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In disaggregated memory systems, hypervisors face challenges in efficiently managing and reclaiming memory pages due to frequent allocation and reclamation needs, especially when virtual machines use Nested Page Tables and modify physical memory, leading to performance overhead.

Innovation Solution

The method involves allocating memory pages to virtual machines in preset units, comparing the address range of page frames to be returned with a preset page size, and removing address space mappings accordingly. This includes removing mappings for contiguous page frames when the return range is equal to or greater than the preset size, and for each page frame when the range is less than the preset size, using a reverse map (rmap) data structure.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If memory pages are allocated to virtual machines in fine granularity for disaggregated memory management, then memory utilization efficiency is improved, but hypervisor performance deteriorates due to frequent allocation and reclamation operations

Engineering Contradiction:

Improvememory utilization efficiencyVSAvoidhypervisor performance overhead

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments memory management into two distinct modes: fine-grained per-page management for disaggregated memory operations and coarse-grained unit-based management for hypervisor operations. This segmentation allows the system to maintain fine-grained control where needed while reducing overhead in hypervisor operations by batching page reclamation into unit-based operations, thus resolving the contradiction between memory utilization efficiency and hypervisor performance.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a memory management intermediary layer that translates between fine-grained page-level requests from the disaggregated memory system and coarse-grained unit-based operations for the hypervisor. This intermediary batches multiple page reclamation operations into unit-based operations, reducing the frequency of hypervisor interventions and improving overall system performance while maintaining fine-grained memory utilization.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If balloon driver technology is used to return memory from virtual machines, then memory reclamation capability is improved, but performance overhead increases significantly

Engineering Contradiction:

Improvememory reclamation capabilityVSAvoidperformance overhead

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent replaces the complex balloon driver mechanism with simpler, disposable unit-based memory reclamation operations. Instead of maintaining a persistent balloon driver within the virtual machine that incurs continuous overhead, the system uses lightweight unit-based operations that can be batched and executed efficiently by the hypervisor, achieving the same memory reclamation capability with significantly reduced performance overhead.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

3Manufacturing precision

If address space mapping is removed for each individual page frame, then memory return granularity is improved, but processing time increases

Engineering Contradiction:

Improvememory return granularityVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent merges multiple individual page frame address space mapping removal operations into batched unit-based operations. By combining adjacent page frames into units and processing them together, the system maintains fine-grained memory return capability (knowing exactly which pages are returned) while reducing the total processing time through batched operations that reduce hypervisor intervention frequency and overhead.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250077261A1Method and apparatus for improving hypervisor performance in memory disaggregation environment

Publication Date: 2025.03.06 ELECTRONICS & TELECOMM RES INST

AI summary

Disclosed herein is a method for improving performance of a hypervisor in a memory disaggregation environment. The method includes allocating memory pages to a virtual machine in preset units, comparing the address range of the page frame to be returned with a preset page size, and removing an address space mapping for the page frame to be returned depending on a result of comparison with the preset page size. Removing the address space mapping comprises removing the address space mapping on the basis of contiguous page frames when the range of the page frame to be returned is equal to or greater than the preset page size.