MORE ON : [htb] strange problems !?
hello again,
I got some just preliminary results .... hope someone can explain them to me... As I already told I have the following config :
egress {
class (30kbps) {sfq};
class (50kbps) {sfq};
class (80kbps) {sfq};
class (30kbps) {sfq};
class (50kbps) {sfq};
...hundreds like this...
class (10kbps,default) {sfq};
}What I got is traffic starvation very often for a period of ~30 sec.. as proposed I've done :
egress {
...hundreds classes...
class (10kbps, prio 7, default) {sfq};
}It seemed a litlle bit better the starvation seemed to be just ~10sec but this time happen much often... then I made a final touch :
egress {
...hundreds classes...
class (55kbps, prio 7, default);
}No "sfq" this time and it seemed much much better now... I have done my tests mostly with :
hping2 host-behind-qos --fast --icmp
fping host1 host2 -l
I made second and third test for very short periods of time so this is not a final "say" of the success...
------------ ?! SO COULD U EXPLAIN ME WHY THIS HAPPEN ?! --------------------------
And most importantly is there a better solution, I don't know what will happen if I want more complex solution but not this flat structure... Or at least give me some directions how to go further and eradicate this "STARVATION" problem..
Can I lower "class (55kbps, prio 7, default);" to "class (10kbps, prio 7, default);" ....
Giving more bandwith to DEFAULT class is not solution but a ANTI-SOLUTION, my idea is not to give all my bandwith to spammers and viruses :")
Expecting explanation :"))
tia
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Hello raptor,
Sunday, October 19, 2003, 2:53:19 AM, you wrote:
At me latest class:
tc filter add dev eth1 parent 1:0 protocol ip prio 200 u32 match ip dst 0.0.0.0/0 police rate 1kbit burst 1k drop flowid 1:0
r> hello again,
r> I got some just preliminary results .... hope someone can explain them to me... r> As I already told I have the following config :
r> egress {
r> class (30kbps) {sfq};
r> class (50kbps) {sfq};
r> class (80kbps) {sfq};
r> class (30kbps) {sfq};
r> class (50kbps) {sfq};
r> ...hundreds like this...
r> class (10kbps,default) {sfq};r> }
r> What I got is traffic starvation very often for a period of ~30 sec.. as proposed I've done :
r> egress {
r> ...hundreds classes...
r> class (10kbps, prio 7, default) {sfq};
r> }
r> It seemed a litlle bit better the starvation seemed to be just ~10sec but this time happen much often... r> then I made a final touch :
r> egress {
r> ...hundreds classes...
r> class (55kbps, prio 7, default);
r> }
r> No "sfq" this time and it seemed much much better now... r> I have done my tests mostly with :
r> hping2 host-behind-qos --fast --icmp
r> fping host1 host2 -l
r> I made second and third test for very short periods of time so this is not a final "say" of the success...
r> ------------ ?! SO COULD U EXPLAIN ME WHY THIS HAPPEN ?! --------------------------
r> And most importantly is there a better solution, I don't know what will happen if I want more complex solution but not this flat structure... r> Or at least give me some directions how to go further and eradicate this "STARVATION" problem..
r> Can I lower "class (55kbps, prio 7, default);" to "class (10kbps, prio 7, default);" ....
r> Giving more bandwith to DEFAULT class is not solution but a ANTI-SOLUTION, my idea is not to give all my bandwith to spammers and viruses :")
r> Expecting explanation :"))
r> tia
r> _______________________________________________
r> LARTC mailing list / [email protected]
r> http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Signature · 6 lines
Best regards, nuclearcat mailto:[email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/