Request dispatch processing method and request dispatch processing system
By introducing multiple candidate request processing queues into the file system, and performing traffic splitting and fine-grained scheduling based on request type and file identifier, the fairness issue between read/write requests and files is resolved, thereby improving the system's throughput efficiency and stability.
Patent Information
- Application Number
- CN202610825955.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-09
- Publication Date
- 2026-07-10
AI Technical Summary
Traditional single-queue distribution mechanisms can lead to read requests being starved by write requests and small files being starved by large files in high-concurrency, mixed-load scenarios, affecting system performance and resource utilization.
Multiple candidate request processing queues are used to divert file processing requests to independent target request processing queues based on request type and file identifier, thereby achieving physical isolation and resource decoupling of read and write requests. Requests are added in a fine-grained manner using file identifiers, and orderly and fair scheduling is carried out through the target request processing queues.
It improves the overall throughput efficiency, response stability, and resource utilization of the file system under mixed workload scenarios such as big data and artificial intelligence training, and avoids the problems of read requests being starved by write requests and small files being starved by large files.
Smart Images

Figure CN122363624A_ABST
