[altq 685] Re: interesting behavior

[Kenjiro Cho <kjc@csl.sony.co.jp>]

Nobuaki Takahashi wrote:
> I encountered an interesting behavior using ALTQ and thought
> if I can get some comments from someone.

[snip]

> First, I tried to transport ATM data stream from A to E. The
> ATM application works perfect in this case. And then, I enabled
> ALTQ at C, which just guarantees bandwidth from A to E. The
> configuration file I used is very simple like this:
> 
> ===========================================================
> interface xl1 bandwidth 100M cbq
> class cbq xl1 root_class NULL pbandwidth 100
> class cbq xl1 def_class root_class borrow pbandwidth 95 default
> class cbq xl1 udp_class def_class pbandwidth 90
> 	filter xl1 udp_class 192.168.3.2 0 0 0 17
> ===========================================================
> (Note: ATM data stream uses UDP and filter shown above reserves
> bandwidth for ATM data stream)
>
> Apparently, this configuration file setup guaranteed bandwidth
> for ATM data stream, where guaranteed bandwidth is at least more 
> than 80Mbps.

Nop.  def_class can also use 95Mbps in this setting.
You should set udp_class as a child of root_class to prevent def_calss
from stealing the bandwidth assigned to udp_class.

> It is expected that ATM application works properly when this
> configuration is setup.
> 
> ... as a matter of fact, the ATM application did not work properly,
> when ALTQ was enabled. According to netstat -w 1, when ALTQ is
> enabled, number of received packets at E becomes larger. I cannot
> understand this behavior.
> 
> What is very interesting is what I have observed next. I had a
> small hope that ATM application will work, when UDP traffic from
> B to F is enabled (probably according to the common sense, it
> does not work properly, since UDP traffic is for torturing ATM
> data stream). And then, when the "interrupting" UDP traffic
> is transported, the ATM application started working.
> 
> I guess the reason ATM application worked is because of the more
> frequent (device?) interruption at C.

It could be.
Have you tried "options HZ=1000" in your kernel config file?

> It would be great if I can get some comment concerning this issue,
> the reason why ATM application did not work when ALTQ was enabled
> and why ATM application started working when interrupting UDP
> data stream is started transporting.

As usual, I need altqstat output for further analysis.

-Kenjiro