Unified Buffer Management for CXL Memory Pooling in DBMS

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing database management systems (DBMS) face limitations in scalability and efficiency due to high latency and lack of cache coherency in disaggregated memory architectures, leading to static resource assignment and high update costs, which constrain scalability and elasticity.

Innovation Solution

Implement a unified memory architecture using Compute Express Link (CXL) for interconnectivity, enabling cache coherency and dynamic memory management, allowing for shared, pooled memory and adaptive resource allocation across multiple memory nodes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If disaggregated memory architecture with RDMA is used, then memory capacity and scalability are improved, but access latency increases and cache coherency is lost

Engineering Contradiction:

Improvememory capacityVSAvoidaccess latency

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The system segments memory management into two layers: pooled memory nodes store data using CXL protocols, while compute nodes maintain local memory caches. This segmentation allows remote memory to be divided into cached portions (for fast access) and pooled portions (for capacity), resolving the latency-capacity tradeoff.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A buffer manager acts as an intermediary between compute nodes and pooled memory nodes. It manages cache coherence, handles memory access requests, and coordinates data synchronization, thereby masking the latency of remote memory access while maintaining the benefits of disaggregated storage.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If static resource assignment is used, then system complexity is reduced, but adaptability to varying workloads deteriorates

Engineering Contradiction:

Improveresource management complexityVSAvoidworkload adaptability

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The buffer manager implements dynamic resource allocation by adjusting cache sizes, buffer allocations, and memory pooling configurations based on real-time workload characteristics. This dynamic adaptation allows the system to optimize performance for different query patterns without requiring complex manual reconfiguration.

Inventive Principle:

Principle #15Dynamics

3Reliability

If explicit DBMS state management is implemented, then data consistency is improved, but update costs and scalability are worsened

Engineering Contradiction:

Improvedata consistencyVSAvoidupdate cost

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The system implements self-service state management through hardware-supported cache coherency protocols and automatic buffer management. The buffer manager autonomously handles consistency maintenance, cache invalidation, and data synchronization without requiring explicit DBMS intervention, thereby reducing update overhead while maintaining reliability.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20260037465A1Unified buffer management in database management systems

Publication Date: 2026.02.05 ACTIAN CORP

AI summary

Aspects described herein relate to accessing, by a compute node of a database management system (DBMS), one or more objects in state information for a database in a unified memory architecture having multiple memory nodes, such as compute express link (CXL), where the accessing includes using direct memory access (DMA) to access memory in at least a portion of the multiple memory nodes, and performing, by the compute node, one or more query operations for the database based on the state information.