Re: CPU Utiliaztion on a ethernet bridge
On Thu, 2003-11-20 at 16:46, Jeff S Wheeler wrote: On Wed, 2003-11-19 at 21:42, Simon Allard wrote: I have replaced NIC's as I thought it might of been the drives also. I moved to the eepro100 cards. Same problem. You should be using NICs with a poll-based driver, as opposed to an interrupt-based driver. This will preempt the kernel less often, and allow it to service the NIC more efficiently. Do you really mean poll-based, or DMA based? Traditionally polling is evil CPU wise... but there could be reasons why polling is better if that is the only thing you are doing. Possibly PC DMA is probably so old and crappy that it's not worth using? -- Donovan Baarda [EMAIL PROTECTED] http://minkirri.apana.org.au/~abo/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
On Thu, 2003-11-20 at 22:34, Donovan Baarda wrote: Do you really mean poll-based, or DMA based? Traditionally polling is evil CPU wise... but there could be reasons why polling is better if that is the only thing you are doing. Possibly PC DMA is probably so old and crappy that it's not worth using? It is my understanding that the modern e1000 driver polls the NIC to find out when new frames are available. You may be correct that it just looks in the DMA rx ring, though; I am a bit out of my league at this point. In either case, the PRO/100 and PRO/1000 cards, using the Intel e100/e1000 drivers, are superb. I suspect the 3c59x driver is not quite so modern, and the kernel is preempted by NIC interrupts frequently when new frames come in under your existing bridge configuration. -- Jeff S Wheeler [EMAIL PROTECTED] Five Elements, Inc. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
On Thu, 2003-11-20 at 16:46, Jeff S Wheeler wrote: On Wed, 2003-11-19 at 21:42, Simon Allard wrote: I have replaced NIC's as I thought it might of been the drives also. I moved to the eepro100 cards. Same problem. You should be using NICs with a poll-based driver, as opposed to an interrupt-based driver. This will preempt the kernel less often, and allow it to service the NIC more efficiently. Do you really mean poll-based, or DMA based? Traditionally polling is evil CPU wise... but there could be reasons why polling is better if that is the only thing you are doing. Possibly PC DMA is probably so old and crappy that it's not worth using? -- Donovan Baarda [EMAIL PROTECTED] http://minkirri.apana.org.au/~abo/
Re: CPU Utiliaztion on a ethernet bridge
On Thu, 2003-11-20 at 22:34, Donovan Baarda wrote: Do you really mean poll-based, or DMA based? Traditionally polling is evil CPU wise... but there could be reasons why polling is better if that is the only thing you are doing. Possibly PC DMA is probably so old and crappy that it's not worth using? It is my understanding that the modern e1000 driver polls the NIC to find out when new frames are available. You may be correct that it just looks in the DMA rx ring, though; I am a bit out of my league at this point. In either case, the PRO/100 and PRO/1000 cards, using the Intel e100/e1000 drivers, are superb. I suspect the 3c59x driver is not quite so modern, and the kernel is preempted by NIC interrupts frequently when new frames come in under your existing bridge configuration. -- Jeff S Wheeler [EMAIL PROTECTED] Five Elements, Inc.
Re: CPU Utiliaztion on a ethernet bridge
Hi! I don't think that I could address your problem, but in Debian there is a package called bridge-utils that allows easy bridge creation. One of my systems uses it for a user-mode-linux installation between tap0 and eth2: - auto tap0 iface tap0 inet manual tunctl_user uml-net auto eth2 iface eth2 inet manual auto br0 iface br0 inet manual bridge_ports eth2 tap0 bridge_stp on As you can notice in my example, the iface method shouldn't be loopback (it isn't the lo device). Try using the manual method if it helps Regards in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
I don't know is it make any sense but I've compiled bridge as integrated, not as module. And I don't state any line in interfaces file about bridged interfaces; Only br0 exists. Also I'm using br0 as external interface. gw11:/etc# gw11:/etc# cat network/interfaces # /etc/network/interfaces -- configuration file for ifup(8), ifdown(8) # The loopback interface # automatically added when upgrading auto lo iface lo inet loopback # # Ayrintili bilgi ve diger parametreler icin: # /usr/share/doc/bridge-utils/README.Debian # auto br0 iface br0 inet static address 212.156.128.21 network 212.156.128.0 netmask 255.255.255.0 broadcast 212.156.128.255 bridge_ports eth0 eth1 bridge_stp on bridge_fd 5 gateway 212.156.128.18 # UNUSED # auto eth2 # iface eth2 inet static # auto eth3 # iface eth3 inet static gw11:/etc# free total used free sharedbuffers cached Mem:257064 211444 45620 0 5276 159648 -/+ buffers/cache: 46520 210544 Swap: 128516 1604 126912 gw11:/etc# gw11:/etc# lsmod Module Size Used byNot tainted shaper 3004 0 (unused) tulip 36928 0 (unused) eepro100 16816 0 (unused) 3c59x 24712 2 gw11:/etc# -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
On Wed, 2003-11-19 at 21:42, Simon Allard wrote: I have replaced NIC's as I thought it might of been the drives also. I moved to the eepro100 cards. Same problem. You should be using NICs with a poll-based driver, as opposed to an interrupt-based driver. This will preempt the kernel less often, and allow it to service the NIC more efficiently. The e1000 driver is excellent in this respect. We run more than 100Mb through a Linux router with a full eBGP table (~127k FIB entries) with no appreciable CPU consumption. The only time the box is substantially taxed is when a BGP peer flaps, in which case zebra consumes a lot of CPU power reconfiguring the FIB. It's a shame that the Linux kernel doesn't make the FIB structures accessible directly via an interface similar to /dev/kmem so zebra could simply mmap(2) it in and make large writes instead of small ioctl(2) calls. -- Jeff S Wheeler [EMAIL PROTECTED] Five Elements, Inc. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
I have replaced NIC's as I thought it might of been the drives also. I moved to the eepro100 cards. Same problem. What kernel are you using? I have tried 2.4.19 and 2.4.22. I'm running the same scenario here on a ppro 200 except that I'm using 2x3c905btxm cards for the briding instead of the one dual card. If I flush all firewall rules i get a throuput of about 95MBit with a cpu-load of constantly under 8%. With all firewall rules i get about 85MBit and the cpu-load is far away from 50%. The bottleneck is the memory timing. Have you tried your setup with two nics as well? I would suggest the problem is the NIC or the driver therefor. Cheers, Michael Simon Allard schrieb: I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message.
Re: CPU Utiliaztion on a ethernet bridge
On Wed, 2003-11-19 at 21:42, Simon Allard wrote: I have replaced NIC's as I thought it might of been the drives also. I moved to the eepro100 cards. Same problem. You should be using NICs with a poll-based driver, as opposed to an interrupt-based driver. This will preempt the kernel less often, and allow it to service the NIC more efficiently. The e1000 driver is excellent in this respect. We run more than 100Mb through a Linux router with a full eBGP table (~127k FIB entries) with no appreciable CPU consumption. The only time the box is substantially taxed is when a BGP peer flaps, in which case zebra consumes a lot of CPU power reconfiguring the FIB. It's a shame that the Linux kernel doesn't make the FIB structures accessible directly via an interface similar to /dev/kmem so zebra could simply mmap(2) it in and make large writes instead of small ioctl(2) calls. -- Jeff S Wheeler [EMAIL PROTECTED] Five Elements, Inc.
CPU Utiliaztion on a ethernet bridge
I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
Any firewall rules or logging mechanism? El mar, 18-11-2003 a las 15:12, Simon Allard escribió: I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: CPU Utiliaztion on a ethernet bridge
No. None what so ever. ipfiler/iptables modules are not even loaded Its a pretty clean running box. Any firewall rules or logging mechanism? El mar, 18-11-2003 a las 15:12, Simon Allard escribió: I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED] Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
CPU Utiliaztion on a ethernet bridge
I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message.
Re: CPU Utiliaztion on a ethernet bridge
Any firewall rules or logging mechanism? El mar, 18-11-2003 a las 15:12, Simon Allard escribió: I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message.
Re: CPU Utiliaztion on a ethernet bridge
No. None what so ever. ipfiler/iptables modules are not even loaded Its a pretty clean running box. Any firewall rules or logging mechanism? El mar, 18-11-2003 a las 15:12, Simon Allard escribió: I have setup a linux box with a 2.4.19 kernel. I am bridging 2 ethernet devices together using 3Com PCI 3c982 Dual Port cards. (3c59x). What I am seeing is that the module itself uses 30% of CPU to handle just 10mbit both ways (20mbit total). ~6000ps total. From what I have read on the bridge homepages I should be able to run this on a 486 as CPU has nothing to do with it. What am I doing wrong? bridge:~# brctl show bridge name bridge id STP enabled interfaces br0 8000.000475c9a6f9 yes eth1 eth2 bridge:~# lsmod Module Size Used byNot tainted limiter13064 0 (unused) bridge 16748 1 3c59x 25512 2 sis900 12388 1 bridge:~# cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file should contain the names of kernel modules that are # to be loaded at boot time, one per line. Comments begin with # a #, and everything on the line after them are ignored. sis900 3c59x options=4,4,4,4 full_duplex=1,1,1,1 max_interrupt_work=1 bridge limiter in /etc/network/interfaces auto eth1 iface eth1 inet loopback auto eth2 iface eth2 inet loopback # Bridge Interface for eth1 + eth2 auto br0 iface br0 inet loopback pre-up brctl addbr br0 up brctl addif br0 eth1 up brctl addif br0 eth2 up brctl stp br0 on down brctl delif br0 eth1 down brctl delif br0 eth2 post-down brctl delbr br0 Does anyone have any ideas off the top of your head what could be causing this or be able to point me in the right direction for some documentation relating to this problem. Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED] Simon Allard (Senior Tool Monkey) IHUG Ph (09) 358-5067 Email: [EMAIL PROTECTED] I'm out of my mind right now, but feel free to leave a message.
