Tiered Storage System I/O Distribution Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Tiered storage subsystems face challenges in designing a tier configuration that meets performance requirements due to the non-proportional relationship between the number of storage media and performance, making it difficult to anticipate and satisfy performance needs.
Innovation Solution
A storage system that counts I/O accesses, creates a cumulative curve of I/O distribution, and allocates RAID groups starting from higher performance tiers based on this distribution to determine the optimal tier configuration, ensuring that performance and capacity requirements are met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If dynamic tiering technique is adopted to automatically allocate pages to appropriate storage media according to access frequency, then storage media are efficiently utilized according to performance needs, but the number of storage media is not simply proportional to performance making it difficult to design in advance the tier configuration capable of satisfying required performance
Solution Approach 1:
The patent applies preliminary action by creating a cumulative curve of I/O distribution before allocating RAID groups. The management device counts I/O accesses, generates the cumulative curve showing I/O distribution characteristics, and uses this pre-computed information to determine the optimal tier configuration. This allows the tier configuration to be designed in advance based on actual I/O patterns rather than making allocations after the fact.
Solution Approach 2:
The patent implements feedback by using the cumulative curve of I/O distribution as feedback information to guide RAID group allocation. The management device continuously monitors I/O accesses, updates the cumulative curve, and adjusts tier configuration decisions based on this feedback. This closed-loop approach ensures that the tier configuration adapts to actual performance requirements while remaining designable in advance.
2Power
If multiple storage media with different performance levels are used to satisfy performance requirements, then performance can be improved, but the non-proportional relationship between number of storage media and performance makes performance prediction difficult
Solution Approach 1:
The patent creates the cumulative curve of I/O distribution in advance, which serves as a predictive tool. By analyzing I/O access patterns and generating this cumulative curve before finalizing the tier configuration, the system can predict the performance characteristics of different storage media combinations and select the appropriate configuration to meet performance requirements.
Solution Approach 2:
The patent changes the parameter representation from simple storage media count to a cumulative distribution curve that captures the actual I/O access patterns. This transformation allows for more accurate performance prediction by considering how I/O requests are distributed across different performance tiers, rather than assuming a simple proportional relationship between media count and performance.
Data Source
AI summary
In tiered storage subsystems in which pages are automatically allocated to appropriate storage media based on the access frequency in page units, since the number of storage media is not simply proportional to the performance, it was difficult to design in advance a tier configuration satisfying the required performance. According to the present invention, a cumulative curve of I/O distribution is created based on a result of measurement of I/O accesses performed to the storage subsystem, and RAID groups (RG) are allocated sequentially in order from RGs belonging to tiers having higher performances to the cumulative curve of I/O distribution. When either a performance limitation value or a capacity of the RG exceeds the cumulative curve of I/O distribution, a subsequent RG is allocated, and the process is repeated so as to compute the optimum tier configuration.