Network Interface Memory Pool Migration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In distributed software systems, data access performance is hindered by the distance and congestion of memory pools, leading to increased latency and network hops, which can violate service level agreements (SLAs) and reduce resource utilization in data centers.

Innovation Solution

A network interface device selects an optimal memory pool based on bandwidth, physical distance, and access frequency to migrate data, using predictive indicators to anticipate congestion and reduce network hops, and notifies the software stack to update memory management schemes, allowing for read-only copies and redirecting memory access requests during migration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If data is stored in remote memory pools in distributed systems, then memory capacity is increased, but data access latency increases and performance decreases

Engineering Contradiction:
Improvememory capacityVSAvoiddata access latency
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The system performs preliminary actions by predicting future data access patterns using machine learning models and proactively migrating data to closer memory pools before actual access occurs. This anticipatory approach reduces latency by ensuring data is already in optimal location when needed, rather than waiting for access requests to trigger migration.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements data copying by creating duplicate copies of frequently accessed data in both remote and local memory pools simultaneously. This allows the system to maintain large memory capacity through remote storage while providing fast access through local copies, effectively resolving the contradiction between capacity and access speed.

Inventive Principle:
Principle #26Copying

2Speed

If data is migrated frequently to optimize access paths, then data access performance improves, but network bandwidth consumption increases

Engineering Contradiction:
Improvedata access speedVSAvoidnetwork bandwidth consumption
Core Design Contradiction:
SpeedVSLoss of energy

Solution Approach 1:

The system employs feedback mechanisms by continuously monitoring actual data access patterns, network conditions, and memory pool performance metrics. This real-time feedback allows the machine learning models to adjust migration decisions dynamically, migrating data only when it will genuinely improve performance and avoiding unnecessary migrations that would waste network bandwidth.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent changes key parameters by using multiple predictive indicators (access frequency, recency, predicted future access) instead of relying on single metrics. This multi-parameter approach allows the system to make more accurate migration decisions, improving data access speed while minimizing unnecessary network traffic by only migrating data that meets multiple criteria.

Inventive Principle:
Principle #35Parameter changes

3Loss of time

If predictive data migration is implemented, then data access latency is reduced, but system complexity increases

Engineering Contradiction:
Improvedata access latencyVSAvoidsystem complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary layer consisting of machine learning models and prediction algorithms that sit between the data access requests and the physical memory migration operations. This intermediary intelligently predicts which data should be migrated and when, simplifying the overall system by centralizing the decision-making logic and avoiding the need for complex coordination between multiple components.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system achieves universality by using a unified machine learning framework that handles multiple functions: predicting data access patterns, determining optimal migration timing, selecting target memory pools, and evaluating migration priority. This multi-functional approach reduces system complexity by consolidating what could be multiple separate complex subsystems into a single versatile prediction engine.

Inventive Principle:
Principle #6Universality (Multi-functionality)

4Speed

If data is kept in local memory pools, then access speed is improved, but available memory capacity is reduced

Engineering Contradiction:
Improvedata access speedVSAvoidmemory capacity
Core Design Contradiction:
SpeedVSQuantity of substance

Solution Approach 1:

The patent applies the copying principle by maintaining duplicate data copies in both local and remote memory pools. Frequently accessed data is copied to local pools for fast access, while the original or additional copies remain in remote pools to preserve total memory capacity. This copying strategy allows the system to simultaneously achieve fast local access and large overall capacity.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system implements dynamics by making memory pool allocation flexible and adaptive rather than static. Data can dynamically move between local and remote pools based on changing access patterns, allowing the system to optimize the balance between local capacity and remote capacity over time. This dynamic allocation ensures that local pools contain the right data for fast access while remote pools provide additional capacity as needed.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4020208A1Memory pool data placement technologies
Publication Date: 2022.06.29 INTEL CORP
  • EP4020208A1 patent drawingFigure 1
  • EP4020208A1 patent drawingFigure 2
  • EP4020208A1 patent drawingFigure 3

AI summary

Examples described herein relate to a network interface device, when operational, configured to: select data of a region of addressable memory addresses to migrate from a first memory pool to a second memory pool to lower a transit time of the data of the region of addressable memory addresses to a computing platform. In some examples, selecting data of a region of addressable memory addresses to migrate from a first memory pool to a second memory pool is based at least, in part, on one or more of: (a) memory bandwidth used to access the data; (b) latency to access the data from the first memory pool by the computing platform; (c) number of accesses to the data over a window of time by the computing platform; (d) number of accesses to the data over a window of time by other computing platforms over a window of time; (e) historic congestion to and/or from one or more memory pools accessible to the computing platform; and/or (f) number of different computing platforms that access the data.