Hi! If I use HTB3 at 100M environment, does it give any improvement if I set HZ=1024?( I''m running an Intel PIII machine). Can you give me a short answer? Thanks in advance. Regards, yangrunhua
Don''t set HZ to 1024 on intel - it a bad idea it produce many problem for other parts of linux which use HZ=100 HTB work fine on 100Mbit without any changes If you use machine with low resources try RED or GRED but its classless 20.06.2002 9:53:50, "yangrunhua" <yangrunhua@njupt.edu.cn> wrote:>Hi! >If I use HTB3 at 100M environment, does it give any improvement if I set >HZ=1024?( I''m running an Intel PIII machine). Can you give me a short >answer? Thanks in advance. >Regards, yangrunhua > >_______________________________________________ >LARTC mailing list / LARTC@mailman.ds9a.nl >http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/ >----------------------------------- mailto:alexey_talikov@texlab.com.uz BR Alexey Talikov FORTEK -----------------------------------
On Thu, Jun 20, 2002 at 10:45:11AM +0500, Alexey Talikov wrote:> Don''t set HZ to 1024 on intel - it a bad idea > it produce many problem for other parts of linux which use HZ=100Can you elaborate? I know of only the ''top'' problem but most other things work well. If you raise HZ you will get more precise shaping and less bursty traffic. Regards, bert -- http://www.PowerDNS.com Versatile DNS Software & Services http://www.tk the dot in .tk http://lartc.org Linux Advanced Routing & Traffic Control HOWTO
I can''t answer you completely ( this question for kernel maintainers ) :( But as I know this value used for IRQ (see arch/i386/kernel/irq.c) different kernel timers (arch/i386/kernel/time.c) for SMP (multiprocessor) in APIC functions ASPI functions almost in all drivers even in PCI bus and ATM subsystem and many many other I have no idea about after-effects of change HZ ! I don''t know about possibility to change this value only for traffic control functions ! I think change this value for Ix86 architecture not so simple ! If somebody have idea about this, share it with me ! 23.06.2002 14:39:59, bert hubert <ahu@ds9a.nl> wrote:>Can you elaborate? I know of only the ''top'' problem but most other things >work well. >If you raise HZ you will get more precise shaping and less bursty traffic. >Regards, >bert >http://www.PowerDNS.com Versatile DNS Software & Services >http://www.tk the dot in .tk >http://lartc.org Linux Advanced Routing & Traffic Control HOWTO----------------------------------- mailto:alexey_talikov@texlab.com.uz BR Alexey Talikov FORTEK -----------------------------------
I change it regulary. I didn''t experienced any problems - only lockups on slow systems which can''t handle it. There should be no problem regarding the change. devik On Mon, 24 Jun 2002, Alexey Talikov wrote:> I can''t answer you completely ( this question for kernel maintainers ) :( > But as I know this value used for IRQ (see arch/i386/kernel/irq.c) > different kernel timers (arch/i386/kernel/time.c) for SMP (multiprocessor) > in APIC functions ASPI functions almost in all drivers even in PCI bus > and ATM subsystem and many many other > > I have no idea about after-effects of change HZ ! > I don''t know about possibility to change this value only for traffic control functions ! > > I think change this value for Ix86 architecture not so simple ! > > If somebody have idea about this, share it with me ! > > 23.06.2002 14:39:59, bert hubert <ahu@ds9a.nl> wrote: > >Can you elaborate? I know of only the ''top'' problem but most other things > >work well. > >If you raise HZ you will get more precise shaping and less bursty traffic. > >Regards, > >bert > >http://www.PowerDNS.com Versatile DNS Software & Services > >http://www.tk the dot in .tk > >http://lartc.org Linux Advanced Routing & Traffic Control HOWTO > ----------------------------------- > mailto:alexey_talikov@texlab.com.uz > BR > Alexey Talikov > FORTEK > ----------------------------------- > > > _______________________________________________ > LARTC mailing list / LARTC@mailman.ds9a.nl > http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/ > >
IMHO :) HZ defines the heart beat of the Linux system in terms of ''jiffies''. Increasing HZ will increase the number of interrupts on the system (e.g. timer interrupt) and hence might not be useful for typical desktops, since handling these system interrupts will not leave much CPU time for applications. But for a server which usually does only one thing (e.g. enforcing QoS), this increase in the interrupts should not be a problem. I configure my servers with HZ=1024 regularly in include/net/pkt_sched.h without any problems. ciao, Amit On Mon, 24 Jun 2002, Martin Devera wrote:> I change it regulary. I didn''t experienced any problems - only > lockups on slow systems which can''t handle it. > There should be no problem regarding the change. > devik > > On Mon, 24 Jun 2002, Alexey Talikov wrote: > > > I can''t answer you completely ( this question for kernel maintainers ) :( > > But as I know this value used for IRQ (see arch/i386/kernel/irq.c) > > different kernel timers (arch/i386/kernel/time.c) for SMP (multiprocessor) > > in APIC functions ASPI functions almost in all drivers even in PCI bus > > and ATM subsystem and many many other > > > > I have no idea about after-effects of change HZ ! > > I don''t know about possibility to change this value only for traffic control functions ! > > > > I think change this value for Ix86 architecture not so simple ! > > > > If somebody have idea about this, share it with me ! > >-- I''m an angel!!! Honest! The horns are just there to hold the halo up straight. ^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^ Amit Kucheria EECS Grad. Research Assistant University of Kansas @ Lawrence (R): +1-785-830-8521 ||| (C): +1-785-760-2871 ____________________________________________________
Exactly :) Alphas used to set 1024 as default. It is because Alpha was very low interrupt latency compared to x86. On Mon, 24 Jun 2002, Amit Kucheria wrote:> IMHO :) > > HZ defines the heart beat of the Linux system in terms of ''jiffies''. > Increasing HZ will increase the number of interrupts on the system (e.g. > timer interrupt) and hence might not be useful for typical desktops, since > handling these system interrupts will not leave much CPU time for > applications. > > But for a server which usually does only one thing (e.g. enforcing QoS), > this increase in the interrupts should not be a problem. > > I configure my servers with HZ=1024 regularly in include/net/pkt_sched.h > without any problems. > > ciao, > Amit > > On Mon, 24 Jun 2002, Martin Devera wrote: > > > I change it regulary. I didn''t experienced any problems - only > > lockups on slow systems which can''t handle it. > > There should be no problem regarding the change. > > devik > > > > On Mon, 24 Jun 2002, Alexey Talikov wrote: > > > > > I can''t answer you completely ( this question for kernel maintainers ) :( > > > But as I know this value used for IRQ (see arch/i386/kernel/irq.c) > > > different kernel timers (arch/i386/kernel/time.c) for SMP (multiprocessor) > > > in APIC functions ASPI functions almost in all drivers even in PCI bus > > > and ATM subsystem and many many other > > > > > > I have no idea about after-effects of change HZ ! > > > I don''t know about possibility to change this value only for traffic control functions ! > > > > > > I think change this value for Ix86 architecture not so simple ! > > > > > > If somebody have idea about this, share it with me ! > > > > > -- > I''m an angel!!! Honest! > The horns are just there to hold the halo up straight. > ^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^-^ > Amit Kucheria > EECS Grad. Research Assistant > University of Kansas @ Lawrence > (R): +1-785-830-8521 ||| (C): +1-785-760-2871 > ____________________________________________________ > > >