Buffers for lossless and lossy traffic are managed differently. Lossy traffic buffers are managed by the egress ports and lossless traffic buffers by the ingress ports. Lossy traffic require a large size buffer to absorb bursts of traffic and which would enable the switch/router to manage these infrequent bursts without resorting to dropping frames.
For lossless traffic, the switch/router can indicate to the sender to stop sending more traffic when congestion is encountered. However, during the time interval between the switch/router sending a congestion signal and the sender receiving and acting on that signal, some traffic would already have been sent. Any frame that is in transit, after the back-pressure signal is sent and before the sender stops transmission, must be absorbed without any drops. Allocation of a portion of the available buffer exclusively to absorb the above traffic will ensure lossless traffic.
To enable lossless support, 25% (twenty five percent) of the packet buffers can be allocated to the lossless pool at device initialization. Once allocated, this buffer cannot be use for lossy traffic. Also, once allocated, the buffer size cannot be shifted between the lossless and lossy buffer pools. Depending on your requirement, you can elect to enable or disable support for this buffer allocation.
Two QoS hardware profiles are provided to enable/disable support for buffer allocation. Select the lossy hardware profile to disable this feature. If you execute any PFC/Pause related commands, they will throw error messages. Select the lossless profile to enable this feature. PFC/Pause configurations can now be applied to the selected interface.
Use the profile qos command, applied to the QoS Hardware context, to
configure the hardware buffer.