hardware and software queuing

[abdul.qadir5001]

What is hardware and software base queuing? Please need your support. Thanks

[Giuseppe Larosa]

Hello @abdul.qadir5001 ,

the hardware queue is actually implemented in hardware and it is just a FIFO  ( First In , First Out ).

Software based queues are used before sending the packet to the hardware queue ( actually the hardware queue size in packet is reduced when CBWFQ or CBWFQ + LLQ are enabled).

The software based queues provides different treatment for packets classified in different traffic classes ( class maps).

With Modular QoS on routers the class maps can be defined based on DSCP value or even more granular using extended IP access lists.

CBWFQ allows to define up to 64 SW queues and it can use absolute bandwidth values or percentages of interface bandwidth ( usually this is equal to the line speed but it can be changed with the interface bandwidth command)

The mechanism is elastic and unused bandwidth by one traffic class can be used by other classes in proportion of their percentages.

LLQ = Low Latency Queue introduces a queue for time sensitive traffic like VOIP. This queue is served first before the standard queues.

To avoid that LLQ queue takes too much resources the value of bandwidth / percentage is an upper limit . In some cases an explicit policer is implemented.

Hope to help

Giuseppe

[balaji.bandi]

adding to @Giuseppe Larosa 

Does load sharing and load balancing are same things?

In General, Load balancing is more of a dynamic technique of distributing the load across 2 or more links while Load Sharing works by statically splitting the traffic types and then distributing the load across multiple links.

[Joseph W. Doherty]

Basically whether queuing is managed by dedicated hardware, like ASICs, or managed by a general CPU.  The former would be the "hardware" approach.  The latter would be a "software" approach.