Storage Load Verification Using Reproduced I/O Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data migration, it is challenging to reproduce the performance metrics of a migration-destination volume when there is a difference in device models between the migration-source and migration-destination storage systems, making it difficult to verify the influence on shared resources and other volumes, especially in hybrid cloud solutions.
Innovation Solution
A load verification system that includes a verification-purposed performance metrics collector, an expected performance metrics data generator, and an I/O pattern data generator to reproduce expected performance metrics by generating and applying I/O patterns based on collected performance metrics and I/O patterns from multiple storage devices, allowing for efficient load verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If trial and error load application is used to verify performance metrics in migration destination, then verification can be performed, but verification time is excessively long
Solution Approach 1:
The system performs preliminary actions by collecting performance metrics and I/O patterns from the migration source storage system before migration. This pre-collected data is stored and later used to generate I/O pattern data for the migration destination, eliminating the need for time-consuming trial and error verification after migration.
Solution Approach 2:
The system copies performance metrics and I/O patterns from the migration source to create representative I/O pattern data for the migration destination. By copying these characteristics and applying them to the destination storage system, the system can verify performance without needing to perform extensive trial and error testing at the destination.
2Measurement precision
If device model differences exist between migration source and destination, then reproduction of expected performance metrics becomes difficult, but verification is still necessary
Solution Approach 1:
The system changes parameters by collecting multiple types of performance metrics (throughput, latency, IOPS) and I/O patterns from the migration source. These parameter sets are stored and used to generate I/O pattern data that adapts to the migration destination's device model, enabling accurate performance verification despite hardware differences.
Solution Approach 2:
The system achieves universality by creating a database of performance metrics and I/O patterns that can be applied across different device models. The I/O pattern data generator uses this universal data to create verification loads that are applicable to various storage systems regardless of specific hardware differences between source and destination.
3Productivity
If shared resources are used in migration destination storage system, then resource utilization is improved, but performance influence on other volumes must be verified
Solution Approach 1:
The system segments the verification process by separately collecting performance metrics for different volumes and shared resources (processors, ports) from the migration source. This segmented data is stored and used to generate targeted I/O patterns that can verify the influence on specific shared resources and other volumes independently, making the complex verification process manageable.
Data Source
AI summary
There is provided a load verification system that performs load performance verification of a data storage area. The load verification system includes: a verification-purposed performance metrics collector that acquires performance metrics of the data storage area and a volume thereof that are load verification targets; an expected performance metrics data generator that generates performance metrics data that will be a result of a load, expected with respect to the load-verification-target volume; and an I/O pattern data generator that generates input/output pattern data based on which a load is generated that causes generation of performance metrics data whose performance is equivalent to expected performance indicated in the expected performance metrics data. An input/output pattern is reproduced by a reproduction section based on the data generated by the I/O pattern data generator, and performance metrics generated as a result of applying a load to the load-verification-target volume are collected.