How to figure out if low WISP speeds are due to traffic shaping orsomething else

Discussion in 'Wireless Internet' started by Chuck Banshee, Feb 15, 2012.

  1. I pay for a certain number of Mbps from my new WISP provider in the Santa
    Cruz mountains yet I get about half that most of the time. A quarter of
    the time I get even less than half that. The WISP is responsive but only
    when I call him.

    Then the speeds suddenly jump up to near the advertised speeds.

    But, a few days later (sometimes only a day later), they drop down back
    to lousy. I call him. They jump back up. A few days later, they drop back
    down to lousy.

    I can tell all this simply by using the speed test web sites, e.g., --> Point to UC Santa Cruz --> Point to San Jose server --> Point to San Francisco server --> Point to San Francisco server

    I 'think' it's all due to the traffic shaping at the access point (the
    WISP can't traffic shape at my radio/antenna because it's my equipment,
    not his so he doesn't have the login).

    But HOW can I tell WHERE the speed bottleneck is occurring?
    (it could be my equipment, his traffic shaping, his equipment, or
    something else altogether).

    What tests can I run to pinpoint the reason the speeds are so bad at the
    times they're so bad?
    Chuck Banshee, Feb 15, 2012
    1. Advertisements

  2. On your way into the valley, you'll notice that there is a freeway
    called Highway 17. It's a fair analogy of wi-fi, a shared highway. If
    the freeway were empty, I'm sure you could probably drive at the
    maximum speed your vehicle offers. However, there are other users on
    Hwy 17. As more and more of them use the highway, the speeds drop.

    However, the analogy doesn't quite work. Freeway drivers can see each
    other and merge into traffic in an orderly manner. Wi-Fi users are
    essentially blind, and cannot synchronize with other traffic on the
    highway. It's literally like driving blind. When you crash into
    something (packet collision), you back off, and try again. It's the
    collison avoidance part of CSMA/CA.

    When you're at a high altitude, you pickup 2.4GHz junk from
    everywhere. One leaky microwave oven and everything within miles goes
    dead. There are literally dozens of other WISP networks, including
    one that I'm working on which you'll probably see (HSMM-MESH) from
    your location. The higher you are, the more you'll hear.

    However, it isn't just your end that might be a problem. The WISP
    central access point might have an omni antenna and/or a better view
    of the world. Interference from other users at the WISP end might be
    causing problems.
    Do you want him to call you instead?
    Well, pretend you're driving down Hwy 17 blindfolded. You can tell
    that you're not going very fast but you have no clue why. What do you
    do? Well, you try to measure the other traffic on the freeway, which
    should give you a fair idea of what speed you could expect. Lots of
    other vehicles means plenty of collisions and slow speeds.

    Start here:
    InSSIDer might be a good start without spending any money. The
    problem is that it shows only wi-fi networks, and does not show
    non-wi-fi sources of interference (RF video links, telemetry, etc).
    However, to do a thorough job, you really need a real spectrum
    analyzer and proper antenna (big dish). Interpreting the results are

    Incidentally, the last time I did some sniffing for a customer (last
    week), the interference source was one of the new and disgusting 40MHz
    wide MIMO routers, that hogs the entire 2.4GHz band. The neighbors
    were streaming video with one, over a distance of about 3ft, but
    clobbering most of the neighborhood because the access point was
    located in a window.
    Ping. Err... fping. Ping the IP address of the WISP wireless access
    point. That limits the results to only you and the WISP. Use fping
    instead of Windoze ping so you can get sequence numbers. Linux ping
    will also work. If it looks like this:
    C:\ZIP\fping>fping -c
    Pinging with 32 bytes of data every 1000 ms:
    Reply[1] from bytes=32 time=1.1 ms TTL=64
    Reply[2] from bytes=32 time=1.1 ms TTL=64
    Reply[3] from bytes=32 time=1.1 ms TTL=64
    Reply[4] from bytes=32 time=1.0 ms TTL=64
    Reply[5] from bytes=32 time=1.1 ms TTL=64
    Reply[6] from bytes=32 time=1.1 ms TTL=64
    where the latency doesn't change much and there are no timeouts or
    lost packets, you're doing fine.

    However, if it looks like this:
    C:\ZIP\fping>fping -c
    Reply[1] from bytes=32 time=155 ms TTL=53
    Reply[2] from bytes=32 time=5.5 ms TTL=53
    Reply[3] from bytes=32 time=15.1 ms TTL=53
    Reply[4] from bytes=32 time=61.9 ms TTL=53
    Reply[5] from bytes=32 time=5.8 ms TTL=53
    Reply[6] from bytes=32 time=177 ms TTL=53
    Reply[7] from request timed out
    Reply[8] from bytes=32 time=16.7 ms TTL=53
    you have a problem. From the latency, the minimum time is probably
    the ideal latency. In this example, it would be about 5msec. Higher
    number indicate retransmissions. The more retransmission, the slower
    your downloads will appear.

    You can also monitor the connection speed of the wireless. It can be
    different in each direction. If the access point is set to variable
    speeds, slower speeds indicate interference or a bad path.

    Phone call... gone.
    Jeff Liebermann, Feb 15, 2012
    1. Advertisements

  3. Chuck Banshee

    miso Guest

    I'm drawing a blank on the term, but isn't there a parameter to request
    longer packets? The first few search results are talking around the
    notion of packet length, but not indicating how to set it.

    My point being if you ask for long packets, maybe you can hog the
    network more.

    The alternative to wisp is satellite, which is far worse due to latency.
    I like satellite internet. It is a good source of free dishes when the
    owner fecal matter cans the gear. I haven't found a good use for the BUC
    that comes with them. I just scrap the LNB affair and put on a new one.
    miso, Feb 15, 2012
  4. Bigger packets in ping or in 802.11? They're different.

    802.11 is nothing more than ethernet encapsulated inside 802.11
    frames. For the details as to packet/frame size, see the 2nd posting:
    If you try to send smaller than 1500 byte ethernet packets, 802.11
    will adjust the frame size to either aggregate, or fragment the
    contents. You can can control the maximum frame size in the advanced
    wireless setting on your wireless router. One reason you might want
    smaller frames is to deal with repetative impulse interference issues.
    Smaller frames have a better chance of getting past interference, than
    larger frames. Thruput will suffer with smaller frames, but at least
    some data will make it through the interference.
    Yep. The idea behind using small packets is to not bring the network
    to a standstill while you're testing. Ping offers no control over the
    size of the packet. For fping, it's
    fping target_ip_address -c -s data_size_bytes
    65500 is the largest allowed value. Note that this does not magically
    create jumbo frames.
    Yeah. I know how it works. I save the old Wild_Blue or HughesNet
    systems for the next desperate owner. We have some local areas that
    can't get DSL, cable, or WISP service, leaving satellite as the only
    suitable last resort. The recent contract settlement with Comcast
    requires Comcast to provide service into some of these areas. I
    expect to see many more dishes and radios at the local recycler
    Jeff Liebermann, Feb 16, 2012
  5. Chuck Banshee

    miso Guest

    I meant longer packets for 802.11 use, not ping. Of course, if the
    connection is flaky, you may just end up getting more packets

    The only way I see the dumped satellite internet gear is from the
    "mountain folk" that have jobs in the flats. These dishes end up in the
    free section of Craigslist. You can do passable FTA on those elliptical
    dishes. You need to replace the LNB holder, and hacksaw a bit to get it
    to fit. But the dishes I've seen are nice Channel Master (Andrews)
    molded fiberglass.

    I've never tried to do wifi over the dish, but that could be an
    alternative for our person with the WISP issues. In theory, a biquad can
    paint the dish.

    Viasat supposed is going to have a next generation high speed internet
    out this year. Still it is hard to fight the speed of light. I know
    someone who has satellite in his motorhome. He skypes his wife, but you
    have to use the word "over" so the comms are only one way.
    miso, Feb 16, 2012
  6. Methinks someone or another (hi, Jeff) pointed out that this
    is (was? [a]) exactly the technique used by Apple for their
    "interference robustness" with their Airports...

    [a] not sure if it's still a functional option.
    danny burstein, Feb 16, 2012
  7. It's part of the 802.11 spec. I'm too lazy to lookup the chapter and
    verse. If the access point is wi-fi compliant, the feature will be
    found in the wireless router or access point "advanced" wireless
    settings. Over the years, I've experimented with the fragmentation
    (and CTS/RTS flow control) mechanism trying to determine if they do
    anything useful. They don't. The problems with fragmentation is that
    it's a performance hit and only works with certain types of
    interference, such as impulse noise and 60/120Hz microwave oven
    bursts. CTS/RTS flow is problematic because of the "hidden
    transmitter" problem, where the client radio cannot hear the
    interfering radio, but the access point can hear both.

    In my never humble opinion, slowing down the wireless speed, and
    fragmenting packets, is not a good solution for interference problem.
    What I do is INCREASE the connection speed to some higher fixed value.
    The logic is that it takes less air time to send a high speed packet,
    than it does to send a low speed packet. I think a 1Mbit/sec packet
    is about 125msec long, while the same amount of data at 54Mbits/sec is
    about 2.5msec[1]. In a noise infested environment, a 125msec long
    packet is sure to get clobbered by a noise burst. A 2.5Msec packet is
    much less likely if it can sneak in between the bursts.

    Some detail:
    Microwave oven interference:

    [1] My guess because I'm too lazy to look it up in the O'Reilly
    802.11 book.
    Jeff Liebermann, Feb 16, 2012
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.