I was donig some testing with Altq v2.2. The testbed is shown as below:
__ _ | |_______
| _| IXIA |____ |
| | |________| | |
loop1 | | | |
| | loop2 | |
| | _________ | |
| |_ | |_ | |
|_____| Router |_____|
IXIA is a traffic generator and measurement sending two loops of traffic flows into ALTQ router at same time. The link is 10Base/T. Two loops are identical traffic of 68 bytes. When two loops traffic loads were increased to around 80% of channel rate, loss rate of one loop (loop1) was increased to 10%, but another's loss rate was just less than 1%. It means two loops behavor quite differently, however the configurations are identical.
We did the same testing on a Normal FreeBSD router in which ALTQ was stop. We found this scenario disappeared. Two loops behavor very similarly. In the case of 80% traffic load at both loops. the loss rates are both less than 1%.
In the router is a FreeBSD v3.3 box, in which the static routing is used. the altq config is below:
#outgoing interface of loop1
interface fxp0 bandwidth 10M cpq
class cbq fxp0 root_class NULL pbandwidth 100
class cbq fxp0 EF root_class pbandwidth 100
filter fxp0 EF 0 0 0 0 0 tos 0xe0 tosmask 0xe0
#outgoing interface of loop2
interface fxp3 bandwidth 10M cpq
class cbq fxp3 root_class NULL pbandwidth 100
class cbq fxp3 EF root_class pbandwidth 100
filter fxp3 EF 0 0 0 0 0 tos 0xe0 tosmask 0xe0
Did anyone have similar experience before? Any comments are highly appreciated.