Alix speedtest capping out at around 8.5 Mbps recently? [SOLVED]



  • Hi,

    I'm totally confused but what I'm seeing in recent speed tests involving Alix boxes. Hopefully someone smarter than me can shed some light on this.

    At home I have an Alix2d3 box running pfSense 2.0-RC1 (i386) built on Thu Mar 24 18:49:39 EDT 2011. It is connected to a Shaw internet connection rated at 15 Mbps down and 1 Mbps up, not including PowerBoost, which usually gives me as high as 28 Mbps in speed tests.

    At work I have an Alix 2d13 box running pfSense 2.0-RC1 (i386) built on Thu Mar 17 08:29:17 EDT 2011. It is connected to a Shaw internet connection rated at 15 Mbps down and 5 Mbps up, not including PowerBoost, which usually gives me as high as 28 Mbps in speed tests.

    Notably, both internet connections are in different cities.

    Up until a few weeks ago (don't know exactly when because I have been upgrading the pfSense firmware frequently over the last few weeks), I had been consistently getting at least 14 Mbps download speed from multiple speedtest sites. More often than not, I had been getting as high at 28 Mbps in speed tests, due to the effects of Shaw PowerBoost.

    Lately, my speedtests have been consistently maxing out at about 8 to 8.5 Mbps on BOTH internet connections. I have tried with and without traffic shaping enabled and I still get the same results. I have tried upgrading the firmware, rebooting, etc, with no success.

    If I hook up say a WRT54GL router running Tomato firmware, I get my usual 14 Mbps to 28 Mbps speed.

    Is it possible that a recent version of pfSense could have introduced a bug or a default setting that could possibly be responsible for this speed loss? Perhaps a new version of the vr (Via?) network driver?

    I'm pulling my hair out here trying to understand what's happening. Any help would be appreciated. Thanks!



  • Are you running any packages?



  • @LostInIgnorance:

    Are you running any packages?

    Only blinkled and RRD summary.



  • Forgot to say that it doesn't matter what time of day I do the speedtest, I still get only 8 to 8.5 Mbps download.



  • Just to check, while you are running the speed test, what does memory, swap and cpu usage show on your boxes? If they top out there might be a problem. do you recall which snapshot of 2.0 was the last one working? Try that one again…



  • @p0ddie:

    Just to check, while you are running the speed test, what does memory, swap and cpu usage show on your boxes?

    Watching the dashboard and top during several speedtest runs on the router at work, CPU usage seems to average 10% and memory usage about 35%. CPU spiked a couple of time to as high as 30% but quickly went back down to 10%. This is no different than I saw several weeks ago when speedtests were fine. Not 100% sure where to find swap usage, BTW.

    @p0ddie:

    do you recall which snapshot of 2.0 was the last one working? Try that one again…

    I wish I did. It might have been a late February snapshot or early March one. It could have been as long ago as the end of December, though. In the future, I'll get into the habit of doing a speedtest after every firmware update.

    I've been perusing the git commit comments for clues but haven't found any yet. There's so many commits in the last few weeks!

    If someone else has an Alix box running a late March build of pfSense on a fast internet connection (i.e. 15 Mbps or more download speed), could you please run a speedtest? Thanks!



  • I'm on time warner road runner, using Alix, and the WAN traffic meter on PFSense said just shy of 15Mb/sec on a speedtest.net run
    They said 30Mb/sec. So my system is running OK on Ver2 and Alix.

    2.0-RC1 (i386)
    built on Fri Mar 18 22:06:27 EDT 2011



  • @pmb1010:

    I'm on time warner road runner, using Alix, and the WAN traffic meter on PFSense said just shy of 15Mb/sec on a speedtest.net run
    They said 30Mb/sec. So my system is running OK on Ver2 and Alix.

    2.0-RC1 (i386)
    built on Fri Mar 18 22:06:27 EDT 2011

    pmb, thanks for that, it was helpful and pointed me in the right direction.

    Okay, I fixed the problem on my home connection. What tipped me off was that 8.5 Mbps what close to 10 Mbps. I figured that perhaps the WAN wire speed was mistakenly negotiated at 10 Mbps instead of 100 Mbps. Sure enough, when I ran ifconfig, I saw this line for the WAN interface:

    media: Ethernet autoselect (10baseT/UTP <full-duplex>)</full-duplex>
    

    I'm not sure why it was negotiated at that speed after my last firmware upgrade. I've done dozens of firmware upgrades in the last several weeks and it was always negotiated at 100 Mbps.

    The fix was to power cycle the "el cheapo" unmanaged Linksys switch that sits between my cable modem and my pfSense Alix router. Never seen that before but you learn something new every day.

    Here is the relevant output of ifconfig after power cycling the switch:

    media: Ethernet autoselect (100baseTX <full-duplex>)</full-duplex>
    

    Tonight I'm getting speeds of 20 to 22 Mbps, in prime time baby!

    I haven't tried fixing the work connection yet. I assume that power cycling the DMZ switch there will work. I'll give it a try and report back here.

    Hope this helps someone else.



  • after having too many linksys units fail, go flakey, I gave up on them.
    I moved on to small HP units, these are great, highly configurable and simply work.
    http://www.amazon.com/HP-J9559A-ABA-Procurve-1410-8G/dp/B003N1ZTC2/ref=pd_sim_e_1



  • @pmb1010:

    after having too many linksys units fail, go flakey, I gave up on them.

    Agreed. It was what I had lying around though.

    @pmb1010:

    I moved on to small HP units, these are great, highly configurable and simply work.
    http://www.amazon.com/HP-J9559A-ABA-Procurve-1410-8G/dp/B003N1ZTC2/ref=pd_sim_e_1

    Highly configurable? It does not seem to be a managed switch. What can you configure on it?



  • Take a look at the 1810 series, not the 1410.  I have an older 1800-8G that has served me well and would not have any concerns buying another.



  • yeah my mistake. I picked the wrong model. It is the 1810G.
    Jumbo frames, port mirroring, VLANs, minor stuff.



  • @cyboc:

    Okay, I fixed the problem on my home connection. The fix was to power cycle the "el cheapo" unmanaged Linksys switch that sits between my cable modem and my pfSense Alix router.

    I haven't tried fixing the work connection yet. I assume that power cycling the DMZ switch there will work. I'll give it a try and report back here.

    As for the work connection speed, the problem was that I was running the tests on a VirtualBox VM. I know what you are thinking but I've never had problems with speedtests on this VM before. Perhaps it was the last VirtualBox update that messed up the speedtest. Anyway, I tried the speedtest on a few physical machines in the office and all of them got between 14 and 14.5 Mbps. In future, I will always try to use a physical machine for speedtests. Stupid me.



  • @cyboc:

    Up until a few weeks ago…I had been consistently getting at least 14 Mbps download speed from multiple speedtest sites. More often than not, I had been getting as high at 28 Mbps in speed tests, due to the effects of Shaw PowerBoost.

    FYI, for any Shaw users out there, please note that despite what Shaw's website currently says, the Server Connect package does NOT include PowerBoost at this time. Shaw called me yesterday to apologize and to say that their website is incorrect and will be updated soon. That is why my download speed tests at work are topping out at about 14 Mbps after switching from the Internet Extreme package to the Server Connect package. More details here.


Locked