Distributed tgt communication optimization method, device and apparatus and storage medium

Abstract

The invention discloses a distributed tgt communication optimization method, device and apparatus and a storage medium, and the method comprises the steps: creating three queues or linked lists for asending thread, and respectively and correspondingly establishing data, disconnection data and message sending data; when the main thread of the tgt calls a corresponding function, constructing a corresponding message type, adding the message type into a corresponding queue or chain table, and sending a reminding character to the sending thread through a pipeline; and after the sending thread receives the reminding character, taking out data from the corresponding queue or linked list through an asynchronous event callback mechanism to execute corresponding operation. Connection establishment,connection disconnection and message sending functions are separated. Connection and event sending are asynchronously processed. A message communication mode is optimized in the mode that the threaddata are maintained through the queue or the linked list, the pressure of the main thread can be reduced, the communication performance is improved, and transmission of large cluster big data is met.

Description

technical field [0001] The present invention relates to the technical field of distributed storage, in particular to a distributed tgt communication optimization method, device, equipment and storage medium. Background technique [0002] iSCSI mainly uses TCP / IP technology to make the storage device (RBD) side through the iSCSI target (iSCSI target) function into a server side that can provide disks, and then through the iSCSI initiator (iSCSI initialization user) function. Clients using iSCSI targets can be mounted, so that disk applications can be performed through the iSCSI protocol. In other words, the iSCSI architecture mainly divides the storage device and the host used into two parts, which are: 1. iSCSI target, which is the device that stores disks or RAID on the storage device side. Currently, the Linux host can also be simulated as an iSCSI target , the purpose is to provide disks used by other hosts; 2. iSCSI initiator, that is, the client that can use iSCSItarge...

AI Technical Summary

Problems solved by technology

If the cluster scale is large, the number of established links will increase exponentially, and the type of data sent and the size of the data packet will also increase sharply. If a message system for sending and receiving is simply built, there will be problems in data maintenance and pressure bearing. It is a big challenge, and doing too many operations on the tgt main thread may cause the system to crash or block

[0004] At present, the message communication between distributed tgt uses a simple message sending and receiving method, such as figure 1 As shown, the thread pool is built on the receiving end, but the establishment and detection of the connection are all in the tgt main thread, which is very prone to exceptions. At the same time, the thread pool does not manage its own links. When the number of links is large, the thread pool Link data maintenance in each thread may go wrong

Images