Now that CoS is crystal clear (ahem...) let's push on to the Bandwidth control feature of the 728TS, which has a much simpler interface. Figure 14 shows the Bandwidth summary page, which contains a link for each port. Clicking a link brings up the settings screen for that port (see Figure 15).

Figure 14: QoS Bandwidth controls
(click image to enlarge)

Unlike CoS, there are Bandwidth controls for both Egress and Ingress traffic. (Note that both "egress" and "ingress" from the point of view of the switch, not the device plugged into a switch port, so you may need to reverse your thinking when setting the controls.) I was glad to see both controls, which gives you the flexibility needed to control either heavy inbound or outbound bandwidth hogging.

Figure 15: QoS Bandwidth controls

You can enter the limits in either kbps or Mbps, and the browser will automatically fill in the other field with the appropriate value. (As a side note, the SW Administration manual contains a few errors in its Bandwidth feature documentation: it refers to kBytes/sec and describes non-existent "committed burst size" controls.)

Note that unless you go over to the Security > Traffic > Storm Control screen and disable Broadcast Control before trying to enable Ingress Bandwidth control, you'll get an error. Fortunately, the error is something like "PORT 1/e1: Storm control is enabled..", since this information isn't provided in the SW Administration guide.

I used the IxChariot setup to check out how well Bandwidth control worked. This time I used one pair of computers and ran two throughput.scr scripts simultaneously, one transmitting traffic (to test Ingress limiting), the other receiving (to test Egress limiting).

I set 50 Mbps Ingress and Egress limits (51200 kbps was the actual value) and then enabled and disabled them during the run in this order:

Figure 16: Bandwidth control test
(click image to enlarge)

The results, as shown in Figure 16, are quite interesting - and certainly not what I expected. The upshot is that egress bandwidth limiting seems to give you throughput somewhat near what you program, while ingress limiting performance is so inaccurate as to appear broken.

I should also note that I'd occasionally get the following when trying to change ingress or egress bandwidth values:

"e="cosparams_shaper_max_burst"
||rlHostParam> ||rlHostParam> ||rlHostParam>