Hybrid switching method for load-balanced unicast services

Abstract

The invention discloses a hybrid switching method for load-balanced unicast services. The method belongs to the technical field of communication. The method comprises the following steps: an input level module obtains a destination port address of a packet; obtaining the destination module address by performing and operation on the bit information of the destination port address; A packet is stored in a buffer queue of a destination module, and a scheduling request is initiated. Receiving a feedback matching scheduling result; The intermediate module receives the scheduling request; Accordingto the number of unprocessed scheduling requests and iterative matching algorithm, the matching scheduling results are obtained and sent. Sending a packet to an output stage module according to a matching scheduling result; The output stage module dispatches a queue head packet in a first-in first-out mode and sends the queue head packet to an output line card. It solves the problem that it is difficult to implement single multicast service simultaneously without increasing a large number of management and storage resources and reducing queue blocking during enqueuing and scheduling. At the same time, it realizes the effect of reducing queue blocking in queue entry and scheduling of unicast multicast services, and does not need to increase a large number of management and storage resources.

Description

technical field [0001] The embodiment of the present invention relates to the technical field of communication, and in particular to a load-balanced single-multicast service hybrid switching method. Background technique [0002] A multi-level switching network is usually composed of multiple small and medium-sized single-level switching units, and has the characteristics of scalability and high flexibility, such as a Clos-type multi-level switching network. Multi-stage switching networks have been increasingly used in the design of high-performance, large-capacity, and multi-port switches. [0003] When a multi-level switching network handles multicast services and single-multicast mixed services, the traditional methods include non-blocking enqueue, Hash mapping enqueue, and complete fan-out enqueue. In the non-blocking queuing method, in order to achieve no head-of-queue blocking effect, the number of queues required is large (2 N -1 queue, N is the number of switching n...

AI Technical Summary

Problems solved by technology

In the non-blocking queuing method, in order to achieve no head-of-queue blocking effect, the number of queues required is large (2 N -1 queue, N is the number of switching network ports); as the number of switching network ports increases, the required non-blocking queue data will increase exponentially, which will take up a lot of logical resources for management and scheduling

In the Hash mapping enqueue method, set a small number of queues (far less than 2 N -1 queue), the address of the multicast group is mapped according to certain rules, such as taking the remainder after modulo the number of queues, and the obtained result is used as the queue number for the multicast group to enter the team; although the number of queues is reduced, it is increased The mapping operation is performed, and there is still a certain degree of head-of-line blocking effect in the queue

In the complete fan-out enqueue method, the arriving multicast packets are completely copied and fan-out according to the destination port at the first time, and the business is scheduled and managed according to the mature unicast scheduling algorithm; however, it is easy to cause the memory of the first-level module to be fanned out. The phenomenon that the replicated group is quickly occupied, and the utilization rate of the storage unit is significantly reduced

[0004] It can be seen that the traditional processing method is difficult to reduce the congestion at the head of the queue when the multicast service is enqueued and scheduled, and does not need to increase a large amount of management and storage resources.

Images