Dynamic Data Migration for Storage Performance Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hierarchical storage systems face challenges in efficiently predicting and moving data from high-load regions between SSDs and HDDs due to static parameter settings, leading to suboptimal performance enhancements.
Innovation Solution
An information processing apparatus that dynamically monitors access logs to identify access concentration regions, predicts future access patterns, and proactively moves data from HDDs to SSDs using transition speed and growth time calculations to optimize storage performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is moved to SSD based on daily access frequency counting, then frequently accessed data is arranged in high-speed storage, but the system cannot respond quickly to short-term IO concentration patterns that last only minutes to tens of minutes
Solution Approach 1:
The patent implements dynamic data arrangement by continuously monitoring IO requests and adjusting data placement between SSD and HDD based on real-time access patterns. The system transitions from static daily-based movement to dynamic, continuous adaptation that responds to changing workload characteristics, enabling quick response to short-term IO concentration while maintaining efficient use of SSD capacity.
Solution Approach 2:
The system employs feedback mechanisms by monitoring IO request patterns and using this information to dynamically adjust data placement decisions. The monitoring unit continuously tracks access patterns, and the movement control unit uses this feedback to determine when and what data to move between storage devices, creating a closed-loop system that adapts to changing conditions in real-time.
2Speed
If the system monitors and moves data regions frequently to respond to IO concentration, then response time to workload changes is reduced, but the complexity of the storage control system increases
Solution Approach 1:
The patent divides the storage system into distinct functional units: a monitoring unit that tracks IO patterns, a determination unit that analyzes the data and decides when movement is needed, and a movement control unit that executes the actual data transfer. This segmentation allows each component to perform its specific function efficiently, reducing overall system complexity while maintaining fast response capabilities.
Solution Approach 2:
The system performs preliminary monitoring and analysis of IO patterns before actual data movement is required. By continuously tracking access patterns and pre-identifying potential IO concentration regions, the system prepares in advance, enabling rapid response when workload changes occur without requiring complex real-time decision-making during the actual transfer.
3Productivity
If data movement is triggered by real-time IO concentration detection, then storage performance is optimized for current workload, but the system may incur excessive data transfer overhead for transient access patterns
Solution Approach 1:
The patent implements a threshold-based approach where data movement is triggered only when IO concentration exceeds predetermined criteria. The determination unit analyzes monitoring data and decides whether the access pattern warrants actual data movement or if it represents transient noise. This partial action approach avoids unnecessary transfers while ensuring that significant workload changes are responded to appropriately.
Data Source
AI summary
An information processing apparatus controlling moving of data stored in a first storage region selected from a plurality of storage regions of a first storage device to a second storage device, the apparatus is configured to execute a collection of logs of accesses, specify the first storage region as an access concentration region based on the logs, move first data stored in the first storage region to the second storage device, specify a transition speed of the access concentration region, specify a first time period until the number of the accesses to the first storage region becomes the first number, at a first time, execute a prediction procedure predicting that the number of accesses to a second region becomes equal to or greater than the first number at a second time after the first time, based on the transition speed and the first time period.