Load Balancing Storage Requests by Path Transfer Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing techniques for distributing I/O operations over multiple paths to a storage system are ineffective when the paths have non-homogeneous transfer speeds, as they often slow down aggregate performance to the speed of the slowest path, failing to optimize bandwidth utilization.
Innovation Solution
A method that determines the transfer speed of each communication path and distributes I/O operations based on the estimated completion time, prioritizing paths with the lowest completion time to maximize aggregate throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If round robin load balancing is used to distribute I/O operations over multiple paths, then simplicity of implementation is improved, but aggregate performance deteriorates when paths have non-homogeneous transfer speeds
Solution Approach 1:
The patent changes the load balancing parameter from simple sequential distribution (round robin) to distribution based on path transfer speeds. The system dynamically determines the transfer speed of each communication path and uses these speed parameters to weight the distribution of I/O operations, thereby optimizing aggregate performance while maintaining implementation feasibility through integrated speed determination and weighted distribution logic.
2Productivity
If I/O operations are distributed over multiple paths with non-homogeneous speeds, then bandwidth utilization can be improved, but complexity of load balancing increases
Solution Approach 1:
The system performs self-service by automatically determining the transfer speed of each communication path through integrated testing mechanisms and using these determined speeds to dynamically adjust load distribution. This eliminates the need for external configuration or manual parameter setting, allowing the system to autonomously optimize bandwidth utilization across heterogeneous paths while managing the complexity internally.
3Productivity
If paths with different transfer speeds are used in parallel, then total throughput can be increased, but completion time uniformity deteriorates
Solution Approach 1:
The patent applies local quality by assigning different distribution weights to different communication paths based on their individual transfer speeds. Instead of uniform distribution, each path receives a proportion of I/O operations相匹配 its performance capability, ensuring that faster paths handle more traffic while slower paths handle less, thereby optimizing total throughput while maintaining acceptable completion time characteristics across the system.
Data Source
AI summary
Methods and systems for distributing I/O requests from a computer system to a storage system over multiple paths having non-homogeneous transfer speeds. A large I/O request is partitioned into a plurality of smaller I/O operations between a computer system and a storage system. The I/O operations are distributed over a plurality of communication paths coupling the computer system and the storage system. The distribution attempts to balance the load over the multiple paths based on the transfer speed of each path. In one embodiment, operations are distributed only to the highest speed paths presently operable. In another embodiment, an estimated completion time for an operation is determined for each path and a path with the lowest estimated completion time is selected to transfer the next operation.