Maximum HSMDA Queue Depth
Since RED slope discards are done based on the current queue depth before allowing a packet
into the queue, a queue may consume buffers beyond the configured MBS value based on the
size of the packet. The maximum number of buffers is based on the high slopes end-buffer
parameter plus the maximum packet size less one buffer. Once the slopes end-buffer is
reached or exceeded, all other packets reaching the queue and associated with the slope will
be discarded. When scheduling removes packets from the queue, the queue depth will
decrease, eventually lowering the depth below the end-buffer threshold.
Control Plane HSMDA RED Slope Policy Management
RED slope configuration is managed by defining up to 1,024 named HSMDA slope policies
on the chassis and mapping the queue to a specific policy name. Each slope policy contains
configuration information for the high-priority slope and the Low-Priority slope. HSMDA
Slope policies differ from standard slope policies in that they do not support the time-average-
factor feature used to manage the weighting utilization of the buffer space the standard slopes
are managing. HSMDA queue slopes operate based on instantaneous queue utilization and do
not maintain a weighted utilization value.
RED slopes used to derive the discard probability based on the current depth of the queue.
Quality of Service Guide
•
The random number is multiplied by the inverse slope value and then added to the
slope Start-Buffer to derive the random fill depth which is the number of buffers that
need to be full in order for the slope discard threshold to cross the random number.
•
If the slope fill depth is equal to or greater than the random fill depth, the packet is
discarded.
•
The packet still may be discarded if a buffer pool associated with the queue has
decremented to the discard point or if the free buffer list is exhausted.
High Scale Ethernet MDA Capabilities
Figure 50
demonstrates the high and low priority
777