Dynamic Memory Page Migration Granularity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Managing data migration between heterogeneous computer memories with different access characteristics, such as latency and capacity, is complex and inefficient due to high overhead from frequent migrations and inefficient use of memory resources.

Innovation Solution

Implementing a dynamic migration granularity technique that adjusts based on application characteristics, using a history-based or profiling-based approach to determine the optimal number of memory pages to migrate, thereby optimizing data placement and reducing migration overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If data is migrated frequently between heterogeneous memories to optimize access characteristics, then memory performance is improved, but migration overhead increases

Engineering Contradiction:

Improvememory access speedVSAvoidmigration overhead time

Core Design Contradiction:

SpeedVSLoss of time

Solution Approach 1:

The patent implements dynamic migration granularity that adjusts the number of pages migrated based on workload characteristics and access patterns. The system transitions from static to dynamic migration strategies, adapting granularity levels (e.g., 1 page, 2 pages, 4 pages, 8 pages) according to real-time conditions, thereby optimizing the balance between migration frequency and overhead.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of migration granularity dynamically. By adjusting the granularity parameter based on application behavior and memory access patterns, the system reduces unnecessary fine-grained migrations while maintaining performance benefits, thus lowering overall migration overhead while preserving speed improvements.

Inventive Principle:

Principle #35Parameter changes

2Loss of time

If migration granularity is increased to reduce migration frequency, then migration overhead is reduced, but memory resource utilization efficiency decreases

Engineering Contradiction:

Improvemigration overhead timeVSAvoidmemory resource utilization efficiency

Core Design Contradiction:

Loss of timeVSProductivity

Solution Approach 1:

The system dynamically adjusts migration granularity based on workload characteristics. For workloads with sequential access patterns or large contiguous memory needs, coarser granularity (e.g., 4 or 8 pages) is used to reduce overhead. For random access patterns or cache-sized workloads, finer granularity (e.g., 1 or 2 pages) maintains efficiency. This dynamic adaptation resolves the contradiction between overhead reduction and efficiency maintenance.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

Different migration granularities are applied to different workloads and memory regions based on their specific characteristics. The system identifies local patterns in memory access behavior and applies appropriate granularity levels locally, rather than using a uniform approach, thereby maintaining efficiency while reducing overall overhead.

Inventive Principle:

Principle #3Local quality

3Quantity of substance

If heterogeneous memories with different capacities are used to balance load, then memory capacity utilization is improved, but access latency varies unpredictably

Engineering Contradiction:

Improvememory capacity utilizationVSAvoidaccess latency

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The system performs preliminary profiling of application memory access patterns before migration decisions are made. By understanding the workload characteristics in advance (e.g., cache size requirements, access patterns), the system pre-determines appropriate migration granularities and targets, thereby avoiding unpredictable latency variations while achieving balanced capacity utilization across heterogeneous memories.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms that monitor memory access latency and performance metrics. Based on this feedback, the migration policy adapts to maintain predictable access times while utilizing the full capacity of heterogeneous memory resources. The feedback loop ensures that latency does not become unpredictable despite capacity balancing.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10503658B2Page migration with varying granularity

Publication Date: 2019.12.10 ADVANCED MICRO DEVICES INC

AI summary

The present disclosure is directed to techniques for migrating data between heterogeneous memories in a computing system. More specifically, the techniques involve migrating data between a memory having better access characteristics (e.g., lower latency but greater capacity) and a memory having worse access characteristics (e.g., higher latency but lower capacity). Migrations occur with a variable migration granularity. A migration granularity specifies a number of memory pages, having virtual addresses that are contiguous in virtual address space, that are migrated in a single migration operation. A history-based technique that adjusts migration granularity based on the history of memory utilization by an application is provided. A profiling-based technique that adjusts migration granularity based on a profiling operation is also provided.