Logical CU Allocation for Scale-Out Storage I/O Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In scale-out type storage systems, the allocation of logical CUs and GDEVs across multiple storage systems can lead to deteriorated I/O performance due to bandwidth limitations and lack of direct connections between storage systems, causing inefficient data access and transfer within the virtual storage system.
Innovation Solution
The solution involves dynamically allocating logical CUs and GDEVs to storage systems based on available resources and access frequencies, with the option to migrate logical CUs or add new ones to systems with sufficient bandwidth, ensuring optimal resource utilization and minimizing data transfer within the virtual storage system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If logical CUs and GDEVs are allocated without considering storage system resources and connections, then allocation simplicity is improved, but I/O performance deteriorates due to data transfer across storage systems with narrow bandwidth
Solution Approach 1:
The patent changes the allocation parameters by considering storage system resources (logical CU count, unused capacity) and connection quality (direct connections, bandwidth) as criteria for allocating logical CUs and GDEVs. This ensures that allocations optimize I/O performance by minimizing data transfer across storage systems with narrow bandwidth while maintaining manageable allocation processes.
Solution Approach 2:
The patent performs preliminary allocation planning by evaluating storage system resources and connection characteristics before allocating logical CUs and GDEVs. This preliminary action prevents performance deterioration by ensuring that high-access-frequency logical CUs and their corresponding GDEVs are allocated to the same storage system with sufficient bandwidth, avoiding the need for later performance corrections.
2Adaptability or versatility
If logical CUs are allocated to multiple storage systems for load distribution, then resource utilization is improved, but data access performance deteriorates when logical CUs and GDEVs are separated across different storage systems
Solution Approach 1:
The patent applies local quality by allocating logical CUs and GDEVs to specific storage systems based on their local characteristics (resource availability, connection quality, access frequency). High-access-frequency logical CUs are locally allocated with their GDEVs to the same storage system to ensure fast data access, while low-access-frequency logical CUs can be distributed to balance resource utilization across multiple storage systems.
Solution Approach 2:
The patent changes allocation parameters by considering access frequency as a key factor. Logical CUs with high access frequency are allocated to storage systems with sufficient resources and direct connections to their GDEVs, while logical CUs with low access frequency can be allocated more flexibly to balance load. This parameter-based allocation maintains both resource utilization and data access performance.
3Quantity of substance
If storage systems are added to increase capacity and scale, then storage capacity is improved, but system complexity increases making resource allocation more difficult
Solution Approach 1:
The patent manages system complexity by introducing clear allocation parameters (storage system resources, connection quality, access frequency) that can be systematically evaluated even as the number of storage systems increases. This parameter-based approach provides a scalable framework for allocating logical CUs and GDEVs across expanding storage infrastructures without proportionally increasing allocation complexity.
Solution Approach 2:
The patent enables storage systems to effectively 'self-select' for allocations by evaluating their own resource availability and connection characteristics. This self-service mechanism allows the system to autonomously determine optimal allocation destinations based on current state, reducing the complexity of centralized allocation management as the system scales.
4Reliability
If logical paths are established to multiple storage systems for redundancy, then system reliability is improved, but bandwidth limitations cause I/O performance deterioration
Solution Approach 1:
The patent applies local quality by establishing logical paths to multiple storage systems with different qualities. High-access-frequency logical CUs are allocated with direct logical paths to their GDEVs on the same storage system for optimal performance, while redundancy is provided through alternative paths. Low-access-frequency logical CUs can utilize multiple storage systems for redundancy without significantly impacting overall I/O performance.
Data Source
AI summary
In a scale-out type storage in which multiple physical storage systems are provided collectively as a single virtual storage system, a logical path is established between the host computer and the virtual storage system so that input/output performance of the storage is not deteriorated, wherein during allocation of a volume to the virtual storage system, if a logical control unit (logical CU) establishing a logical path to a volume is unallocated, a logical CU and a volume is generated to a storage system having either a small number of allocated logical CUs or a small amount of used storage capacity. On the other hand, if there is a storage system having a logical CU already allocated thereto, a volume is generated in that storage system.