[Click] Click questions

[Runting]

Hi, Eddie, thank you so much for your help.
I already solved the second problem, the problem is that pull-to-push
elements require a lot of scheduling overhead. And also RoundRobinSched does
a for loop to check if each queue is empty in each scheduling cycle, and
this turned out to be expensive too where I had > 1000 queues, and most of
the time the queues are all empty. So I got rid of the pull-to-push element,
and did something else that had the same effect, and I wrote my own
RoundRobin Scheduler that has constant overhead per scheduling cycle.

btw, I don't know if this is specific to the version I downloaded,
LinuxIPLookup code seems to be missing one line.

int
LinuxIPLookup::init_routes(ErrorHandler *errh)
{
  if (_out2dev)
    delete[] _out2dev;
  _out2dev = new net_device * [_nout];
  _out2dev[0] = 0;
  int i;
  for(i = 0; i < _nout; i++){
    net_device *dev = dev_get_by_name(_out2devname[i].cc());
    click_chatter("LinuxIPLookup: dev = %x\n", dev);
        if (dev == 0)
      return errh->error("Cannot find device %s", _out2devname[i].cc());
//the following line is missing
    _out2dev[i] = dev;
  }
  return(0);
}


Thanks again!

Runting
----- Original Message ----- 
From: "Eddie Kohler" <kohler at CS.UCLA.EDU>
To: "Runting" <runting at cs.cmu.edu>
Cc: "Yih-Chun Hu" <yihchun at crhc.uiuc.edu>; "Bryan Parno" <parno at cmu.edu>;
"Adrian Perrig" <adrian at ece.cmu.edu>; <click at amsterdam.lcs.mit.edu>; "Bruce
Maggs" <bmm at cs.cmu.edu>
Sent: Saturday, October 29, 2005 11:47 AM
Subject: Re: [Click] Click questions


> Hi,
>
> I wasn't sure from your description, thanks for the explanation.  But
> the patch definitely adds all the symbols it needs.  If Roman has bad
> luck perhaps you can mail the Emulab patch to the list and we can try
> to combine them.
>
> How about the problematic configuration and expected number of
> queues?  I understand that you have fixed the problem, but in a non-
> modular way -- we would like the modular way to work.
>
> E
>
>
>
> On Oct 28, 2005, at 4:05 PM, Runting wrote:
>
> > Hi, thanks a lot, Eddie.
> >
> > No, of course I didn't only change the source. I recompiled the entire
> > kernel, booted from the new kernel, and then tried to start Click.
> > I just learned today that Roman Chertov has managed to do the same
> > thing for
> > his research, so I am contacting Emulab Operations hoping they can
> > make the
> > disk image public.
> >
> > Thanks a lot for your help.
> >
> >
> > Runting
> >
> >
>