Low throughput



  • Ive now had this PFsense box running for quite a long time, since november I beleive. Im on cable internet, and it bursts up to 16Mb/s or so for quite some time, if im downloading large files, as long as the network traffic is low. To make things sort, It seems Pfsense on my current computer still has serious issues with lack of throughput. Ive tried setting up a second PC (Vmware machine, NT4) on the WAN side, but I havent had any luck.

    I have reinstalled Pfsense, it stopped working/booting after a power faliure… All of my outlets are hooked to the circuit breaker marked "LIGHTS"... Assuming the LIGHTS circuit breaker woudl cut power to the lights, I flipped it to repair a light socket... Anyways.

    After reinstalling Pfsense, it was still performing about the same. http://www.wolfgame.org/crapbox/router_iperf.JPG
    There are only 3 PC's on this network, counting the PFsense box. One 933Mhz Hewlett packard running solaris, which uses next to nothing for bandwidth, considering its not being used for anything. One gaming PC running XP, and Pfsense. Between the solaris box and the PC, I can get close to 80Mb/s, im fairly certain we can rule out any network bottlenecks.

    System information (from status.php):
    AMD K6-500Mhz, SIS530 chipset mATX motherboard

    FreeBSD 6.1-RELEASE-p2 #0: Fri Jun 16 01:03:36 UTC 2006
    sullrich@builder.livebsd.com:/usr/obj.pfSense/usr/src/sys/pfSense.6
    Timecounter "i8254" frequency 1193182 Hz quality 0
    CPU: AMD-K6(tm) 3D processor (501.14-MHz 586-class CPU)
    Origin = "AuthenticAMD"  Id = 0x58c  Stepping = 12
    Features=0x8021bf <fpu,vme,de,pse,tsc,msr,mce,cx8,pge,mmx>AMD Features=0x80000800 <syscall,3dnow>real memory  = 333381632 (317 MB)
    avail memory = 316571648 (301 MB)

    Network:
    10/100 dual port COMPAQ Netelligent server nic, ~1997 I beleive.

    tl0: <compaq 10="" 100="" netelligent="" dual="" port="">port 0xb800-0xb80f irq 10 at device 0.0 on pci1
    miibus0: <mii bus="">on tl0
    nsphy0: <dp83840 10="" 100="" media="" interface="">on miibus0
    nsphy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
    tl1: <compaq 10="" 100="" netelligent="" dual="" port="">port 0xbc00-0xbc0f irq 11 at device 1.0 on pci1
    4 port 10/100 belkin switch, previously was a 24 port managed rackmount switch.

    The compaq dual port NIC was in my gaming PC, and it performed quite well, could nearly max out each port. What exactly should I start looking at, in order to improve performance. I understand this is fairly vintage hardware, but a 500Mhz K6 should be able to do more than 10BaseT speeds.</compaq></dp83840></mii></compaq></syscall,3dnow></fpu,vme,de,pse,tsc,msr,mce,cx8,pge,mmx>



  • Make sure by reviewing your dmesg output that you have no other device using one of the nic's irqs. When you see the slow throughput, do you see your box running at 100% cpu? Also do you use the trafficshaper? Next thing to try would be using different nics for the pfSense box I think.



  • @hoba:

    Make sure by reviewing your dmesg output that you have no other device using one of the nic's irqs. When you see the slow throughput, do you see your box running at 100% cpu? Also do you use the trafficshaper? Next thing to try would be using different nics for the pfSense box I think.

    *The Dual-port NIC is using IRQ 10 for TL0, and IRQ 11 for TL1.
    *When the throughput "peaks" or reaches its highest point, the CPU is near 80-100%
    *No Traffic shaper



  • Try enabling polling mode at system>advanced (not sure if your nics support it though). I guess your nics start to use a lot of interrupt time (view top at the shell when it uses up the CPU).



  • Recent snapshots include vmstat.

    Login to a shell/console and access option 8.  Then run:

    vmstat -i

    Once you feel the box is going slow and it will show you the interrupt rate such as:

    builder6# vmstat -i
    interrupt                          total      rate
    irq1: atkbd0                          6          0
    irq14: ata0                          46          0
    irq28: bge0                    97204652        45
    irq32: amr0                    252741638        117
    cpu0: timer                    22512650        10
    cpu1: timer                    22510768        10
    Total



  • @sullrich:

    Recent snapshots include vmstat.

    Login to a shell/console and access option 8.  Then run:

    vmstat -i

    vmstat: command not found

    Which folder do I need to navigate to? Tried to figure it out through google, no luck.





  • Iperf Results, Polling Disabled
    Vmstat Results, Polling Disabled
    Iperf & Vmstat results, Polling Enabled
    Iperf results to 127.0.0.1 on the 450Mhz PC

    If its only doing 40-50Mbs to 127.0.0.1, I have a feeling the NICs arnt the problem though I could be wrong.

    I just ran this under vmware on my older 1.8Ghz pc (XP SP2 as the host-OS, pfsense installed in vmware). 700-800Mb/s to 127.0.0.1, and 80-95Mbs over Vmwares 10/100 ethernet adapter to my other pc, cpu load pegged at 100%.


Log in to reply