I/O Data Caching Based on Transfer Rate Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing caching methodologies are ineffective in filtering I/O data for remote storage environments, as they do not account for the slower data generation and consumption rates, leading to inefficient caching and bottlenecks due to remote data transfer limitations.
Innovation Solution
A method that filters I/O data based on the generation/consumption rate by calculating the transfer rate of ongoing sequential data streams and caching only when the rate exceeds the throughput of the target storage device, optimizing the use of the caching device by avoiding caching of slower data streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data is cached based on recency, frequency, block sizes, and file types, then local data access performance is improved, but remote data transfer efficiency deteriorates due to bottlenecked connections
Solution Approach 1:
The patent changes the caching decision parameters from traditional metrics (recency, frequency, block size, file type) to include transfer rate comparison. By evaluating whether the data generation/consumption rate exceeds the target storage device throughput, the system adapts caching behavior to the actual data flow characteristics, resolving the contradiction between local access speed and remote transfer efficiency
Solution Approach 2:
The system dynamically adjusts caching decisions based on real-time transfer rate measurements. Rather than using static caching policies, the system continuously monitors data stream characteristics and makes adaptive caching decisions, allowing it to optimize for either local performance or remote efficiency depending on the current data flow conditions
2Speed
If all I/O data is cached to enhance performance, then data retrieval speed is improved, but system resources are wasted on caching low-throughput data streams
Solution Approach 1:
Instead of caching all I/O data universally, the patent applies partial caching action by selectively caching only those data streams where the transfer rate exceeds the target storage device throughput. This partial action approach avoids the energy waste of caching low-throughput data while still providing performance benefits for high-throughput streams
Solution Approach 2:
The patent applies different caching quality levels to different data streams based on their individual characteristics. High-throughput data streams receive full caching treatment for maximum performance, while low-throughput streams receive no caching to avoid resource waste, creating a localized quality approach rather than uniform treatment
3Loss of time
If caching is used for remote storage devices, then local data access is accelerated, but the bottlenecked remote connection creates inefficiency
Solution Approach 1:
The system performs preliminary evaluation of the data stream transfer rate before making caching decisions. By measuring whether the data generation/consumption rate exceeds the target storage device throughput in advance, the system determines whether caching will be beneficial, preventing wasted preliminary caching actions on remote data that would not improve overall performance
Data Source
AI summary
An invention is provided for filtering cached input/output (I/O) data. The invention includes receiving a current I/O transfer. Embodiments of the present invention evaluate whether to filter ongoing data streams once the data stream reaches are particular size threshold. The current I/O transfer is part of an ongoing sequential data stream and the total data transferred as part of the ongoing sequential data stream is greater than the predetermined threshold. The transfer rate for the ongoing sequential data stream then is calculated and a determination is made as to whether the transfer rate is greater than a throughput associated with a target storage device. The current I/O transfer is cached when the transfer rate is greater than the throughput associated with a target storage device, or is not cached when the transfer rate is not greater than the throughput associated with a target storage device.