Storage Server Data Writing Method Using Write-Only Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing write policies in storage systems are inefficient, particularly in cloud workloads, as they fail to differentiate between write-only and ordinary data, leading to unnecessary cache traffic and reduced read performance.
Innovation Solution
A data writing method that categorizes data into write-only and ordinary data based on historical access patterns, using the write-around policy for write-only data and write-through or write-back policies for ordinary data, to minimize cache traffic and optimize cache utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If write-back policy is used to write data to cache first and then asynchronously flush to hard disk drive, then write performance is improved, but cache space is wasted on write-only data that will never be read
Solution Approach 1:
The patent segments data into two categories: write-only data and ordinary data. Write-only data is identified by monitoring access patterns (data written but never read within a threshold time), and is then handled by a dedicated write-around policy that bypasses the cache entirely. Ordinary data continues to use the write-back policy with cache. This segmentation resolves the contradiction by directing write-only data away from the cache, preserving cache space for readable data while maintaining high write performance for both data types.
2Reliability
If write-through policy is used to write data to both cache and hard disk drive simultaneously, then data consistency is improved, but write traffic to cache increases unnecessarily for write-only data
Solution Approach 1:
The patent segments data into write-only and ordinary data based on access pattern analysis. For write-only data, it applies the write-around policy that writes directly to the hard disk drive without involving the cache, thereby eliminating unnecessary write traffic to the cache while maintaining data consistency. For ordinary data, the write-through policy is maintained to ensure data consistency. This selective application resolves the contradiction by removing wasteful cache traffic for write-only data while preserving consistency guarantees for readable data.
3Device complexity
If a fixed write policy is applied to all data, then system simplicity is maintained, but write policy efficiency decreases in diverse cloud workloads
Solution Approach 1:
The patent introduces a dynamic write policy that adapts to different data types based on observed access patterns. The system monitors write and read operations, identifies write-only data through threshold-based analysis, and dynamically applies different write policies (write-around for write-only, write-back for ordinary data). This dynamic adaptation resolves the contradiction by allowing the system to optimize write policy efficiency for diverse cloud workloads while maintaining manageable complexity through automated pattern recognition and policy selection.
Data Source
AI summary
This application discloses a data writing method, device, a storage server and a computer readable storage medium, including: writing, when a write request is received, write data corresponding to the write request to a write buffer; acquiring historical access data of a data block corresponding to to-be-flushed data in the write buffer when a data flushing operation is triggered for the write buffer; determining whether the to-be-flushed data is write-only data based on the historical access data by using a pre-trained classifier; if yes, writing the to-be-flushed data to a hard disk drive; and if no, writing the to-be-flushed data to a cache. The data writing method provided by this application can effectively reduce the traffic of writing dirty data to the cache while reserving more space in the cache for the ordinary data, thereby improving the utilization of the cache space and the read hit rate of the cache.