Flow-aware switch shared cache scheduling method and device

Abstract

The invention discloses a flow-aware switch shared cache scheduling method and device. The method comprises the following steps: in the operation process of the switch, monitoring whether events of data packet queue entering, data packet queue exiting, data packet loss, cache overflow and queue state change of each output port of the switch occur or not in real time; judging that the flow state of the port is light load, heavy load or overload according to the port event; determining the control state of the current port to be a common state, an absorption state or an emptying state based on the port flow state; and based on the port control state, adjusting the cache threshold of the port of the switch, and changing the upper limit of the cache which can be used by the port. Through real-time monitoring of port traffic, different ports can be subjected to differential management according to requirements, so the use efficiency of the shared cache of the switch is effectively improved.

Description

technical field [0001] The present invention relates to the technical field of switch shared cache, in particular to a flow-aware switch shared cache scheduling method and device. Background technique [0002] In computer networking, switch buffers are used to absorb uneven traffic arriving at switch ports. In order to improve the efficiency of cache usage, currently commonly used commercial switches usually use on-chip shared memory (on-chip shared memory). All outgoing ports of the same switch share a unified buffer area, and different ports perform statistical multiplexing on the shared buffer area. Because all ports share the cache, when some ports of the switch are heavily loaded, some ports may occupy the entire cache while other ports cannot use the cache, resulting in unfairness among ports. In order to avoid the above unfairness, it is necessary to The shared cache is managed through the shared cache scheduling policy. [0003] The current mainstream shared cache ...

AI Technical Summary

Problems solved by technology

Due to the lack of port traffic awareness in existing dynamic threshold schemes, long-term traffic that does not need to use cache resources occupies the cache for a long time, resulting in insufficient cache for burst traffic that requires cache resources

Images