Distributed Storage Metadata for Cache Bypass

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional distributed storage systems inefficiently manage data storage in cache and non-cache resources, often storing data in cache even if it is unlikely to be accessed, leading to wasted storage capacity in cache resources.

Innovation Solution

Implementing a distributed storage system where metadata is generated to determine whether data should be stored in cache or non-cache resources, preventing unnecessary storage in cache for replica data that is unlikely to be accessed, and instead storing it in non-cache resources to free up cache space for more frequently accessed data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If data is stored in cache resources to improve access speed, then access time is reduced, but storage capacity is wasted when the data is unlikely to be accessed

Engineering Contradiction:
Improvedata access speedVSAvoidcache storage capacity utilization
Core Design Contradiction:
SpeedVSQuantity of substance

Solution Approach 1:

The system performs preliminary actions by generating metadata that predicts future data access patterns before data is actually accessed. This allows the system to proactively determine which data should be cached, ensuring that cache resources are allocated to data that will likely be accessed, thereby avoiding both premature caching and cache misses

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms by continuously monitoring data access patterns and using this information to update metadata. This feedback loop allows the system to adapt its caching decisions based on actual usage patterns, improving cache hit rates and ensuring that cache capacity is allocated efficiently to frequently accessed data

Inventive Principle:
Principle #23Feedback

2Reliability

If replica data is stored in cache to ensure data availability, then reliability is improved, but cache resources are wasted on data unlikely to be accessed

Engineering Contradiction:
Improvedata availabilityVSAvoidcache storage capacity
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The system applies local quality by differentiating between primary data and replica data in metadata, and by location-specific caching decisions at different nodes. Each node makes localized caching decisions based on its role (primary or replica) and local access patterns, rather than applying a uniform caching strategy across all data

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system performs preliminary actions by generating metadata that identifies replica data before it is stored. This allows the system to proactively prevent replica data from being cached, ensuring that cache resources are reserved for primary data that requires high availability, while replica data is stored in non-cache resources

Inventive Principle:
Principle #10Preliminary action

3Speed

If all received data is stored in cache by default, then data access speed is improved, but storage cost increases due to high-capacity cache requirements

Engineering Contradiction:
Improvedata access speedVSAvoidstorage cost
Core Design Contradiction:
SpeedVSLoss of energy

Solution Approach 1:

The system changes parameters by dynamically adjusting caching behavior based on data characteristics identified in metadata. Rather than using a fixed caching strategy, the system modifies caching parameters (such as cache eligibility, retention time, and priority) based on factors like data type, access frequency, and data importance, thereby optimizing the balance between access speed and storage cost

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3080717B1Cache and non-cache usage in a distributed storage system
Publication Date: 2020.03.04 INTEL CORP
  • EP3080717B1 patent drawingFigure 1
  • EP3080717B1 patent drawingFigure 2
  • EP3080717B1 patent drawingFigure 3

AI summary

According to one configuration, upon receiving data, a respective node in a distributed storage system produces metadata based on the received data. The generated metadata indicates whether or not to bypass storage of the received data in the cache storage resource and store the received data in the non-cache storage resource of the repository. Data storage control logic uses the metadata to control how the received data is stored. A state of the metadata can indicate to prevent storage of the received data in a corresponding cache resource associated with the respective storage node. Thus, the generated metadata can provide guidance to corresponding data storage control logic whether to store the received data in a cache storage resource or non-cache storage resource.