Sequential I/O Workload Detection in Superimposed Storage Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized and container environments, storage systems struggle to detect sequential I/O workloads within a mixed I/O stream, leading to pseudo-random I/O patterns and reduced cache hit rates, as multiple applications' I/Os are consolidated onto a single volume, hindering the ability to optimize read latencies and prefetch data effectively.
Innovation Solution
A method to detect individual sequential I/O workloads within a stream of superimposed workloads by analyzing time dependencies in block accesses, calculating a sequentiality factor for each workload, and optimizing cache strategies based on this analysis to separate and prioritize sequential I/O streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple workloads are consolidated onto a single volume to improve storage capacity utilization, then storage efficiency is improved, but the ability to detect sequential I/O patterns deteriorates
Solution Approach 1:
The patent segments the mixed I/O stream into individual workload streams by analyzing temporal patterns and block access sequences. Each workload's sequential I/O pattern is identified and separated through pattern recognition algorithms that distinguish between different applications' access behaviors, enabling sequential detection even in consolidated volumes.
Solution Approach 2:
The patent introduces an intermediary layer (storage software or controller) that sits between the consolidated volume and the storage backend. This intermediary analyzes I/O patterns, identifies sequential workloads, and applies optimization strategies selectively, allowing sequential I/O detection and optimization without requiring volume separation at the host level.
2Device complexity
If I/O workloads are superimposed on a single I/O queue to simplify storage management, then management complexity is reduced, but cache hit rate deteriorates
Solution Approach 1:
The patent implements dynamic cache management that adapts to different workload patterns within the same I/O queue. The system continuously monitors I/O patterns, identifies sequential workloads in real-time, and dynamically adjusts cache allocation and prefetching strategies, maintaining high cache hit rates despite workload superposition.
Solution Approach 2:
The patent applies different cache optimization strategies to different workloads within the same I/O queue based on their sequentiality characteristics. Sequential workloads receive aggressive prefetching and higher cache priority, while random workloads receive standard handling, optimizing overall cache performance without requiring workload separation.
3Use of energy by moving object
If sequential I/O detection is performed at the volume level without additional processing, then processing overhead is minimized, but detection accuracy deteriorates in virtualized environments
Solution Approach 1:
The patent applies partial processing by focusing computational resources only on identifying sequential patterns rather than analyzing every I/O operation in detail. The system uses lightweight pattern recognition that processes only the necessary temporal and spatial relationships in block accesses, achieving accurate sequential detection with minimal processing overhead.
Data Source
AI summary
In one general embodiment, a computer-implemented method includes detecting individual sequential input/output (I/O) workloads in a stream of superimposed I/O workloads accessing a same physical volume. The detecting is based on a time dependency corresponding to accesses of blocks of the volume. In another general embodiment, a computer-implemented method includes detecting a plurality of sequential input/output (I/O) workloads in an I/O stream of superimposed workloads accessing a same volume, the detecting being based on a time dependency corresponding to accesses of blocks of the volume. A sequentiality factor is calculated for each of the sequential I/O workloads.