If I use a QNX 6 box as an IP gateway for other boxes, is there a
generally accepted method or technique to limit the transfer rate
to specific hosts behind the gateway?
If one host behind the gateway is downloading a large file(s) it
appears to negitivly impact performance to other hosts behind the
gateway. I certainly understand why. My question is, is there anything
that can be done about it?
It would be nicer still if, under circumstances where there was no
contection for the link, that the box demanding the throughput was
actually granted it. But I can live without this.
I was figuring that something could be done with ip-filter, but it
appears that ip-filter would then be responsible for buffering possibly
large amounts of data. Surely I’m not the first one to think that this
would be a nice feature.
–
Bill Caroselli – Q-TPS Consulting
1-(626) 824-7983
qtps@earthlink.net
It is a nice feature.
What you wanted is called “traffic shaping”. Based on certain rules, (where
the
packet from, what port it is going…), packets are forwarded in the decided
manner (1 packet per second, or 20% of abaliable bandwidth). So you can
limit out going ftp traffic, or limit the traffic hitting your port 80,
leave you some
space to telnet into your box 
Unfortunatly ipfilter can’t do shaping. The only thing I know of, is
openBSD’s
dummynet, but that is tied into their stack and is not an easy take to port
it.
Besides, I want to write my own stuff (once I have some time )
-xtang
Bill Caroselli <qtps@earthlink.net> wrote in message
news:b1992u$ofh$1@inn.qnx.com…
If I use a QNX 6 box as an IP gateway for other boxes, is there a
generally accepted method or technique to limit the transfer rate
to specific hosts behind the gateway?
If one host behind the gateway is downloading a large file(s) it
appears to negitivly impact performance to other hosts behind the
gateway. I certainly understand why. My question is, is there anything
that can be done about it?
It would be nicer still if, under circumstances where there was no
contection for the link, that the box demanding the throughput was
actually granted it. But I can live without this.
I was figuring that something could be done with ip-filter, but it
appears that ip-filter would then be responsible for buffering possibly
large amounts of data. Surely I’m not the first one to think that this
would be a nice feature.
–
Bill Caroselli – Q-TPS Consulting
1-(626) 824-7983
qtps@earthlink.net
Xiaodan Tang <xtang@qnx.com> wrote:
It is a nice feature.
What you wanted is called “traffic shaping”. Based on certain rules, (where
the
packet from, what port it is going…), packets are forwarded in the decided
manner (1 packet per second, or 20% of abaliable bandwidth). So you can
limit out going ftp traffic, or limit the traffic hitting your port 80,
leave you some
space to telnet into your box >
Unfortunatly ipfilter can’t do shaping. The only thing I know of, is
openBSD’s
dummynet, but that is tied into their stack and is not an easy take to port
it.
Besides, I want to write my own stuff (once I have some time )
-xtang
Cool. SO I’m not the only one to think of it.
Are you hinting that it will/might one day be a part of the QNX stack?
I’m in no real rush for this one.
I would assume that implementing this would require some very large
buffering to be done somewhere.
Bill Caroselli <> qtps@earthlink.net> > wrote in message
news:b1992u$ofh$> 1@inn.qnx.com> …
If I use a QNX 6 box as an IP gateway for other boxes, is there a
generally accepted method or technique to limit the transfer rate
to specific hosts behind the gateway?
If one host behind the gateway is downloading a large file(s) it
appears to negitivly impact performance to other hosts behind the
gateway. I certainly understand why. My question is, is there anything
that can be done about it?
It would be nicer still if, under circumstances where there was no
contection for the link, that the box demanding the throughput was
actually granted it. But I can live without this.
I was figuring that something could be done with ip-filter, but it
appears that ip-filter would then be responsible for buffering possibly
large amounts of data. Surely I’m not the first one to think that this
would be a nice feature.
–
Bill Caroselli – Q-TPS Consulting
1-(626) 824-7983
qtps@earthlink.net
Xiaodan Tang <xtang@qnx.com> wrote:
Bill Caroselli <> qtps@earthlink.net> > wrote in message
news:b19sf2$e4h$> 3@inn.qnx.com> …
Cool. SO I’m not the only one to think of it.
Are you hinting that it will/might one day be a part of the QNX stack?
I’m in no real rush for this one.
If it implemented, it will be something like IPFilter, which is not part of
the TCPIP stack (but co-exist in io-net world). However, as far as I
know, currently there is no such plan.
I would assume that implementing this would require some very large
buffering to be done somewhere.
Not really. You can choose the action to give a limit of how much packet
to buffer, so that too many packets will result packet lost. Also, things
like TCP will likely have flow control kick in, the sender will detect
longer Round Trip Time, and start sending data “slower”.
-xtang
Gotchya!
I just figured that since TCP is supposed to be “reliable” that the
system should do it’s part not to “damage” the data. I.E. loose packets.
–
Bill Caroselli – Q-TPS Consulting
1-(626) 824-7983
qtps@earthlink.net
Bill Caroselli <qtps@earthlink.net> wrote in message
news:b19sf2$e4h$3@inn.qnx.com…
Cool. SO I’m not the only one to think of it.
Are you hinting that it will/might one day be a part of the QNX stack?
I’m in no real rush for this one.
If it implemented, it will be something like IPFilter, which is not part of
the TCPIP stack (but co-exist in io-net world). However, as far as I
know, currently there is no such plan.
I would assume that implementing this would require some very large
buffering to be done somewhere.
Not really. You can choose the action to give a limit of how much packet
to buffer, so that too many packets will result packet lost. Also, things
like TCP will likely have flow control kick in, the sender will detect
longer Round Trip Time, and start sending data “slower”.
-xtang
Bill Caroselli <> qtps@earthlink.net> > wrote in message
news:b1992u$ofh$> 1@inn.qnx.com> …
If I use a QNX 6 box as an IP gateway for other boxes, is there a
generally accepted method or technique to limit the transfer rate
to specific hosts behind the gateway?
If one host behind the gateway is downloading a large file(s) it
appears to negitivly impact performance to other hosts behind the
gateway. I certainly understand why. My question is, is there
anything
that can be done about it?
It would be nicer still if, under circumstances where there was no
contection for the link, that the box demanding the throughput was
actually granted it. But I can live without this.
I was figuring that something could be done with ip-filter, but it
appears that ip-filter would then be responsible for buffering possibly
large amounts of data. Surely I’m not the first one to think that this
would be a nice feature.
\
Bill Caroselli – Q-TPS Consulting
1-(626) 824-7983
qtps@earthlink.net