Dynamic resource regulation and control method and system for guaranteeing delay sensitive application delay SLO

Abstract

The invention provides a dynamic resource regulation and control method and system for guaranteeing delay sensitive application delay SLO. The method includes: creating a plurality of request queues in a service node of the distributed storage system, wherein different types of requests are located in different queues, thread groups are allocated to the request queues according to logic thread resources and target tail delay requirements of the service nodes, the thread resources are dynamically regulated and controlled in real time, and the thread groups of all the request queues are bound with physical CPU resources; and enabling client to send an application access request to the service node, enabling the service node to store the application access request into a request queue corresponding to the type of the application access request as a current queue according to the type of the application access request, process the application access request by using the thread group allocated to the current queue, and send a processing result to the client. According to the invention, the tail delay of the delay-sensitive application can be ensured to meet the target requirement, and the bandwidth of the batch processing application can be maintained at a relatively high level.

Description

technical field [0001] The invention relates to the technical field of distributed storage systems, in particular to the technical field of ensuring low-tail latency requirements of delay-sensitive applications. Background technique [0002] Distributed storage systems take a typical client-server (Client-Server) model. The request processing process for accessing the distributed storage system is: (1) the application (Application) requests the client to access the distributed storage system; (2) the client sends the request to the server through the network; (3) the server accepts the request from the network And start processing. After the request arrives at the server, it will first enter the queue to queue. If there is an idle thread, it will fetch the request from the queue and process it; (4) After the request is processed on the server side, its response will be sent to the network through the network. Client; (5) After the client receives the response to the request...

AI Technical Summary

Problems solved by technology

[0014] The purpose of the present invention is to solve the problems of resource competition and inflexible thread model in the above-mentioned prior art, which lead to delay-sensitive application tail delays that cannot meet the target requirements and low resource utilization, and propose a window-based method for dynamically adjusting thread resources

Images