[Click] some outputs from peak rate measurement, and questions
Lan Xue
lanxue at soe.ucsc.edu
Sat Aug 14 11:26:00 EDT 2004
Hi,
On Sat, 14 Aug 2004 gcalarco at deis.unibo.it wrote:
> Quoting Lan Xue <lanxue at soe.ucsc.edu>:
>
> > Hi all,
> >
> > Thanks for those who give me suggestions about measuring the peak packet
> > forwarding rates. The click router is installed on a PentiumIII machine
> > with 1.2GHz CPU and 1G memory. Because the for now, the 100Mb network
> > adapters are the bottleneck, I instead use InifiteSource as traffic
> > generator and Counter to collect the number of packets passed. Kernel
> > space
> > provides around 1.1M packets per sec, about twice as much as that provided
> > by user level click.
> >
> > I tried 64Byte, 1kByte and 10KByte packet, but didn't observe any obvious
> > difference in forwarding rates between different packet sizes. Does that
> > imply that there's no upper bound for the number of bits passed through
> > the router per sec, given the number of the packets are not higher than
> > the peak rate?
> >
>
> yes, you are not using the PCI bus at all, that's change
> things a lot...
>
Is that(involving the PCI bus) measurable without using extra machine but
just within the click machine?
Thanks,
Lan
packet size does not matter at the click (software) level, since
> what is passing through the click elements is a packet
> pointer, not the packet itself.
>
> ciao
> giorgio
>
>
> > Next are the handlers printed out by PokeHandlers, running at user-level.
> > It was called in this way: PokeHandlers(read ctr.rate, read ctr.count,
> > read ctr.byte_count, read myqueue.drops, write ctr.reset, wait 1, loop); I
> > assumed that ctr.count should be close to ctr.rate if not exactly the
> > same, since there's no drops of packets from the queue. Also, I thought
> > ctr.count should be the product of ctr.byte_count and size in byte per
> > packet. However, neither is the case as I expected, as you can see from
> > below:
> >
> > ctr.rate:
> > 518795.996
> >
> > ctr.count:
> > 743040
> >
> > ctr.byte_count:
> > 40867200
> >
> > myqueue.drops:
> > 0
> >
> > ... more repeated prints, but with similar values as the above for each
> > handler.
> >
> > Anyone has any ideas/comments/suggestions?
> >
> > Thanks a lot,
> >
> > Lan
> >
> > _______________________________________________
> > click mailing list
> > click at amsterdam.lcs.mit.edu
> > https://amsterdam.lcs.mit.edu/mailman/listinfo/click
> >
>
>
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program
> provided by the DEIS Webmail System.
>
More information about the click
mailing list