Here is what did...
Auto QoS set on the 48 port switches. Any devices on the 24TS connecting to another device on the 24TS works fine. Devices connecting from the 48P to the 24TS with Auto-QoS on the 48P experince severe performance degradation. Below is the global command to fix this.
**** Global Command to Fix the Issue
mls qos queue-set output 1 threshold 2 400 400 100 400
****** Queue Setting Before Fix **********
MID3750Stack#sh mls qos queue-set 1
Queueset: 1
Queue : 1 2 3 4
----------------------------------------------
buffers : 20 20 20 40
threshold1: 100 50 100 100
threshold2: 100 50 100 100
reserved : 50 100 50 50
maximum : 400 400 400 400
********** Queue Setting After Fix *************
MID3750Stack#sh mls qos queue-set 1
Queueset: 1
Queue : 1 2 3 4
----------------------------------------------
buffers : 20 20 20 40
threshold1: 100 400 100 100
threshold2: 100 400 100 100
reserved : 50 100 50 50
maximum : 400 400 400 400
Notice that in the output above that the thresholds for queue 2 by
default are only configured to allow 50% of their buffer to be used
before traffic is dropped since Queue 2 is the lowest priority queue
and services the type of traffic that your switch has been
dropping. We changed this queue to accomodate a higher amount of the
lower priority traffic which thus resolves the issue.
As you can see, there are core differences in the architecture of the
two switches which account for the different behaviors when it comes
to buffer management.
