Re: IPRewriter and ICMP packets

Erik VandeKieft wrote:
> Hmmm. Actually, now that I think about it, I guess I could have been using
> just an IPAddrRewriter rather than an IPRewriter all this time, since I
> never translate any ports. I can always look up my mappings based on just
> the source and destination IP address.
> 
> Since that doesn't look at ports, couldn't that work with ICMP ping
> packets? The ICMPPingRewriter as its written still isn't enough because it
> wants a static NEWSRC and NEWDST; I at least want it to look in the table
> of an IPAddrRewriter.

Hey Erik,

You may be interested in the new IPAddrPairRewriter element (just checked in 
to CVS), which is probably better suited than IPAddrRewriter or IPRewriter for 
your needs.

Eddie

Click polling extensions

Hi,

I was wondering anyone can talk about the Click polling extensions. What
is it about and what device driver(s) need to be tuned to take advantage
of it. And anyone has experience of approximately how much more packets
Click can forward after enabling the polling extensions?

Thanks a lot,

Lan

Re: Click polling extensions

> Hi,

Hi,

> 
> I was wondering anyone can talk about the Click polling extensions. What
> is it about and what device driver(s) need to be tuned to take advantage
> of it. And anyone has experience of approximately how much more packets
> Click can forward after enabling the polling extensions?
> 

If I remember right, there's something about Click polling performance in this paper:
http://www.pdos.lcs.mit.edu/papers/click:tocs00/paper.pdf

It's a little old, but I think it can give you an idea.

Cheers,
--
-------------------------------
Juan Luis Baptiste M.
Linux registered user #119248
http://www.merlinux.org
http://knetworkconf.sf.net

"we're back to the times
 when men were men and
 wrote their own drivers"
        Linus Torvalds
-------------------------------

Re: Click polling extensions

----- Original Message ----- 
From: "Lan Xue" <lanxue <at> soe.ucsc.edu>
To: <click <at> amsterdam.lcs.mit.edu>
Sent: Tuesday, August 03, 2004 7:38 PM
Subject: [Click] Click polling extensions

> Hi,
>
> I was wondering anyone can talk about the Click polling extensions. What
> is it about and what device driver(s) need to be tuned to take advantage
> of it. And anyone has experience of approximately how much more packets
> Click can forward after enabling the polling extensions?

hi,

it dependes a lot on the kind of hardware you have.
Some other examples (in addition to the papers):

- Intel PRO 1000 nic + PCI 1.0 bus  + 1.6 Ghz Pentium III -> 350Kpps
- Intel PRO 1000 nic + PCIX  bus + double 2.8 Ghz Xeon -> 1250Kpps

ciao
giorgio

Ing. Giorgio Calarco
DEIS - Università di Bologna
Viale Risorgimento, 2 40136 Bologna - Italy
Tel: 051 2093776 Fax: 051 2093073
E-mail: gcalarco <at> deis.unibo.it
PGP: http://www3.deis.unibo.it/GCalarco/Giorgio_Calarco.asc

Re: Click polling extensions

Hi,

On Wed, 4 Aug 2004, Giorgio Calarco wrote:

> ----- Original Message -----
> From: "Lan Xue" <lanxue <at> soe.ucsc.edu>
> To: <click <at> amsterdam.lcs.mit.edu>
> Sent: Tuesday, August 03, 2004 7:38 PM
> Subject: [Click] Click polling extensions
>
>
> > Hi,
> >
> > I was wondering anyone can talk about the Click polling extensions. What
> > is it about and what device driver(s) need to be tuned to take advantage
> > of it. And anyone has experience of approximately how much more packets
> > Click can forward after enabling the polling extensions?
>
> hi,
>
> it dependes a lot on the kind of hardware you have.
> Some other examples (in addition to the papers):
>
> - Intel PRO 1000 nic + PCI 1.0 bus  + 1.6 Ghz Pentium III -> 350Kpps
> - Intel PRO 1000 nic + PCIX  bus + double 2.8 Ghz Xeon -> 1250Kpps
>

In the above examples, is it machine to machine test or test within router
machine using InfiniteSource() as traffic generator? If it's machine to
machine test, could you tell what ethernet cards were using(over 1Gbps I

Re: Click polling extensions

> >
> > it dependes a lot on the kind of hardware you have.
> > Some other examples (in addition to the papers):
> >
> > - Intel PRO 1000 nic + PCI 1.0 bus  + 1.6 Ghz Pentium III -> 350Kpps
> > - Intel PRO 1000 nic + PCIX  bus + double 2.8 Ghz Xeon -> 1250Kpps
> >
>
> In the above examples, is it machine to machine test or test within router
> machine using InfiniteSource() as traffic generator?

the first one. TX -> click router -> RX

>If it's machine to
> machine test, could you tell what ethernet cards were using(over 1Gbps I
> guess?)?

The Intel PRO 1000 adapter, which is a 1 gb/s card and one of the few
supporting polling. There is a whole family of cards called PRO1000,
just take a look at the Intel web site for the details and the differences
among them. I felt attracted for the Intel PRO1000 XT Server Adapter,
(it has got a copper interface), but there are many...

>
> What's the packet size for the above examples?

right, I forgot, sorry.
Packet size is 64 byte. If you use longer packets
packet rate (i mean the MLFFR) decreases
(but throughput in bit/s increases till the card

www.pdos.lcs.mit.edu/grid/software.html

Hello,
We are creating an IT Directory at http://www.the-it-resource.com and we would like to include 
your website http://www.pdos.lcs.mit.edu/grid/software.html to our Software category - http://www.the-it-resource.com/Software/.
Please submit your site's information here: http://www.the-it-resource.com/Software/form.html.

A reciprocal link from your site would be appreciated. This is the linking code :
<a href="http://www.the-it-resource.com/Software/">Software resources</a> - 
directory of Software related websites.

Thank you,
Michael Westmore,
webmaster <at> the-it-resource.com.
::text version::

_______________________________________________
click mailing list
click <at> amsterdam.lcs.mit.edu
https://amsterdam.lcs.mit.edu/mailman/listinfo/click

running kernel-level click

Hi,

I have this problem of running click at kernel-level: ./click-install
returns successfully, and /click FS has been mounted. However ping from
one machine to the other which are interconnected by click got
"desintation host unreachable" errors. The router configuration file works
fine with userlevel click.

Please help, thanks,

Lan

Re: peak packet forwarding rate measurement

Hi Eddie,

On Thu, 29 Jul 2004, Eddie Kohler wrote:

> If you are running a kernel configuration, the handlers are available
> in the /click filesystem.
>

Does the content under /click be refreshed for each click uninstall?

> If you are running a user level configuration, you can include a
> ControlSocket element and telnet to the router to see handler values
> (man ControlSocket).
>
> Either way, you can include a PokeHandlers element to print handler
> values periodically (man PokeHandlers).
>

PokeHandlers works fine for userlevel but it doesn't print out(to standard
output) for kernel version. ./click-install CONFIGFILE simply returns
without any output. Although, I can see the values of handlers under
/click. Did I miss something to get the periodical handler print?

Thanks,

Lan

> Eddie
>
> On Jul 29, 2004, at 1:12 PM, Lan Xue wrote:

Re: running kernel-level click

that's ok, use this elements to have packets (not only ICMP ones)
passed to the Linux kernel (so it receives the ICMP request):

http://www.pdos.lcs.mit.edu/click/doc/ToHost.n.html

or vice versa (to allow the Linux kernel to send out packets):
http://www.pdos.lcs.mit.edu/click/doc/FromHost.n.html

otherwise you can also change your click config inserting
the ICMPPingResponder...

ciao
giorgio

Ing. Giorgio Calarco
DEIS - Università di Bologna
Viale Risorgimento, 2 40136 Bologna - Italy
Tel: 051 2093776 Fax: 051 2093053
E-mail: gcalarco <at> deis.unibo.it
PGP: http://www.deis.unibo.it/GCalarco/Giorgio_Calarco.asc

----- Original Message -----
From: "Lan Xue" <lanxue <at> soe.ucsc.edu>
To: <click <at> amsterdam.lcs.mit.edu>
Sent: Saturday, August 07, 2004 3:33 AM
Subject: [Click] running kernel-level click

> Hi,
>
> I have this problem of running click at kernel-level: ./click-install