Page 1 of 1

VERY big difference in CPUs usage on RB1100AHx2

Posted: Sun Feb 19, 2012 9:24 pm
by ManUtd
Hi,

RB1100AHx2, ROS 5.13, latest firmware. About 300/40mbit of traffic and I have very big usage of only one CPU. First 10 ports bridged, ether13 using as uplink. Even, in /profile, it some CPU load on firewall but I don't have ANY filter, nat or mangle rule.

Is there any way to make usages of CPUs almost same ? (tried to manually set IRQs but it didn't help)

Anyone with same issue ?

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Sun Feb 19, 2012 10:45 pm
by frankie
On x86 architecture with Atom D525 CPU same difference. Firewall CPU Load probably means that Connection Tracking is ON.

Image

Load of CPU cores:
Image

RB1100AHx2, ROS 5.13, latest firmware. About 300/40mbit of traffic and I have very big usage of only one CPU. First 10 ports bridged, ether13 using as uplink. Even, in /profile, it some CPU load on firewall but I don't have ANY filter, nat or mangle rule.

Is there any way to make usages of CPUs almost same ? (tried to manually set IRQs but it didn't help)

Anyone with same issue ?

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Thu Feb 23, 2012 5:39 pm
by DobroFenix
I too want to take it, but I'm afraid .....
You pps 70k at 400 megabits
I have a PPP 50k at 50 megabits
What I am doing wrong and whether it will cope with my problems?
NAT (100), queues (30), firewall (30)

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Thu Feb 23, 2012 8:13 pm
by ManUtd
frankie,

No, conn. tracking is OFF. I see that you have same situation even with only 2 ether i-faces. Did you try to update to 5.14?

DobroFenix,

Maybe you have some loops in network, do /tool torch to found out that

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Fri Feb 24, 2012 7:48 am
by normis
this is completely normal. each process can take one core. same process can't be split between cores. so you will never see same usage on two cores.

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Fri Feb 24, 2012 8:53 am
by ManUtd
normis,

thats ok but why cannot select which process using which CPU ? I mean, why i.e. queuing or firewall is not processing by CPU which have less load ?

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Fri Feb 24, 2012 9:04 am
by normis

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Sat Feb 25, 2012 11:07 pm
by DobroFenix
frankie
RPS has helped you?
It has become better?

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Sat Feb 25, 2012 11:36 pm
by w0lt
I dont' see "RPS" as an option when I enter "System Resource" on my RB1100AHx2 (fw: 5.14).

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Sun Feb 26, 2012 1:54 am
by frankie
RPS was always turned on. No change in load under ROS 5.14.
frankie
RPS has helped you?
It has become better?

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Sun Feb 26, 2012 11:50 pm
by ManUtd
RPS was always turned on. No change in load under ROS 5.14.
frankie
RPS has helped you?
It has become better?
Thanks for info frankie.

Re: VERY big difference in CPUs usage on RB1100AHx2

Posted: Mon Feb 27, 2012 4:33 am
by bdennie
Something else to consider is to move your uplink port off ether 13. I have seen this on all RB1100 series routers I have used. Ethers 11,12,13 have higher ping times the ports 1-10. Now the pings only spike up to 15ms but when in any other ports the pings are 1ms consistent. Just noticed what port you were using for uplink an thought I would see if this helps with latency.