Message order preserving method based on serial number mechanism

Abstract

The invention discloses a message order preserving method based on a serial number mechanism, which comprises the following steps that: S1, when asynchronous message sending operation is carried out each time, a sending end packages a current sending serial number into a packet header, and adds 1 to the sending serial number; S2, a receiving end compares the current receiving serial number with the sending serial number carried in the packet header every time the receiving end successfully receives the asynchronous message; S3, after the message requests arriving in sequence are successfully processed, 1 is added to the value of a message serial number generator at the receiving end, and whether elements matched with the current receiving serial number exist in the elements arriving in a out-of-sequence queue or not are determined; and S4, if it is determined that the asynchronous message arrives out of order once, the message buffer area which finishes receiving this time is added into an out-of-order arrival queue, and processing is performed after the subsequent message arrives and the receiving serial number is updated. Concurrency of communication paths can be effectively guaranteed, the problem of message sequences caused by a bottom-layer interconnection framework can be effectively solved, and meanwhile influences of serial number generation and matching on performanceare very low.

Description

technical field [0001] The invention relates to a message sequence preservation method based on a serial number mechanism, which belongs to the technical field of computers. Background technique [0002] The message programming specification based on the message model usually stipulates the order of the multiple communication behaviors of the two parties. Taking the two communications between two processes as an example, the first message sent by the source must be earlier than the one sent by the source. When the second message arrives at the destination address space, if the destination address of the two messages is the same, the final content of the destination address must be written by the second message. [0003] Due to the logic mechanism design of the underlying message components or the design of the interconnection network architecture, the underlying network cannot guarantee the orderly arrival of messages at the software level. On the one hand, the network inter...

AI Technical Summary

Problems solved by technology

However, the shortcomings of this technology are also obvious. First, the above method can only guarantee the message sequence of a single physical channel. Under the condition of multi-channel network, it still cannot guarantee the message sequence of multiple physical channels. Secondly, the above method strictly limits the communication at a certain time. There can only be one message in flight in the channel, and the subsequent messages must wait for the response of the previous message before sending out, resulting in a sharp decline in communication performance, so that the network performance cannot be fully utilized

Images