Re: Slow browsing with cable modem

From: P Gentry (rdgentry1_at_cablelynx.com)
Date: 04/26/04

  • Next message: David Efflandt: "Re: Expanding Wireless coverage by cascading APs"
    Date: 25 Apr 2004 21:02:00 -0700
    
    

    "Stephen Zilliox" <szilliox@charter.net> wrote in message news:<108numtbmmr063c@corp.supernews.com>...
    > > Strange... Have you checked your proxy settings?
    > No proxies set as far as I know. I checked under the Mandrake linux control
    > center.
    > >
    > > Can you ping default GW? (68.113.6.1)
    >
    > [root@stevescomputer stephen]# ping 68.113.6.1
    > PING 68.113.6.1 (68.113.6.1) 56(84) bytes of data.
    > From 10.210.32.1 icmp_seq=26 Packet filtered
    > From 10.210.32.1 icmp_seq=27 Packet filtered
    > From 10.210.32.1 icmp_seq=91 Packet filtered

    Won't even let you ping your GW -- low blow, foul! But you really
    shouldn't ping endlessly this way. Try ping -c4 x.x.x.x
    >
    > --- 68.113.6.1 ping statistics ---
    > 186 packets transmitted, 0 received, +3 errors, 100% packet loss, time
    > 184974ms

    Oops, errors again -- thought we had lost those. An immediate
    ifconfig here could have provided a tad more info re: the errors. It
    _is_ one advantage to sending so many packets.

    > > Traceroute ?
    > >
    > [root@stevescomputer stephen]# traceroute -n 68.113.6.1
    > traceroute to 68.113.6.1 (68.113.6.1), 30 hops max, 38 byte packets
    > 1 * * *
    > 2 * * *
    > 3 * * *
    > 4 * * *
    > 5 * * *
    > 6 * * *
    > 7 * 10.210.32.1 74.893 ms *

    Usually running traceroute to your GW is a wasted effort as you don't
    get anything back that is useful except this "hop" indicator that
    shows a kind of shortest loop. Yours returns in 7, mine in 3 --
    74.9ms vs 9.3ms. Even with the extra hops this is a _long_ time to
    return. Maybe that's why they reject pings to your GW -- it's slow
    too?

    > > Is this the case...?:
    > > Cable modem <----> eth0 in your Mandrake box
    > > No extras?
    > PCI ethernet card in my computer attached to cable modem to cable to
    > Charter. I used DHCP under both Win XP and Linux to try to connect to the
    > web. I did not ding around changing connection parameters unless I was
    > following suggestions from you folks and that was only after I had tried
    > everything I could think of using the Mandrake Control Center. I suspect
    > that there is probably a wide variety of skill levels represented on these
    > groups, but it is difficult to know whose advice to follow.

    We sometimes don't appreciate how confusing so many voices can be --
    especially in the throws of net connection woes. J.M. is bringing you
    along OK -- no damage done and lots of info from you that helps to
    provide leads or eliminate suspects. Yours may be a subtle problem
    that's hard to pin down. Be patient with us if you can as we sift
    through the evidence.

    > I appreciate
    > all the help and advice I get but I do realize I could possibly really muck
    > up my system. One of the first things I tried with this problem was to
    > reinstall the whole Mandrake system. Didn't help. I also moved from
    > Mandrake 9.2 to 10.0 thinking that perhaps there were bugs that had possibly
    > been worked out. Didn't work. When I was on 56k dialup, I spent most all
    > of my time in Linux, but since web access is important to any computer nut I
    > have been forced to spend most of my time in--as you say--Windoze.
    >
    > Steve Z.

    Some notes while I wait for the traceroutes from my end to time out...

    Re: Configuration

    Your last 2 network config IPs look OK to me -- ie., at least they are
    consistent and make sense. GW on same net segment as you and you can
    ping it. Can you ping name servers consistently?

    These (your IP, GW, and name servers),are the basic set of IPs needed
    to get out to the internet. They come from the DHCP server that gives
    you your lease and your IP.

    You'll note that the first 2 name servers are on a different network
    from the last two -- 4 name servers on two geographically separated
    nets indicates someone expects high traffic loads and a need for
    backup servers.

    When ping and traceroute give consistent results to these IPs, your
    network _connectivity_ is not a problem -- neither at your end nor at
    the ISP's. Throughput rate is another matter. Consistent pings stats
    indicate "normal". Slow times and/or dropped packets to these servers
    or GW indicate ISP network problems.

    Heavy net loads will slow DNS lookups especially -- ie., most
    noticable with web surfing. Differences in browser performance
    indicate differences in browsers. Note that Moz does do background
    (look ahead) fetches -- this is known to cause performance degradation
    in some specific network environments. This could be compounded by
    ISP's caching proxies that are not in synch -- even worse if you're
    running Squid or have local (Moz) caching set just wrong. Try
    different caching settings in Moz. Both of these are under Edit -
    Preferences - Advanced - Cache. Also check Mozilla site for settings
    available only with a text editor -- sorry I don't remember which
    support/help/faq page you'll need to find the link.

    Re: Hardware
    Your first post showed errors in ifconfig output, the last did not.
    (Alas, the ping above did.) The errors were few in number but were
    based on "corrupt" frames -- ie., bad signals on the wire. The low
    number of errors and their intermittent appearance usually result in
    much hair pulling -- that's why I have so little left ;-( Larger
    numbers usually indicate auto-negotiation failure between the nic and
    CM. Check with mii-tool -v for link status and speed as a double
    check. man mii-tool for other data collection methods.

    Can you run (probably via a DOS sys floppy) the Netgear diagnostics?
    http://kbserver.netgear.com/kb_web_files/N100159.asp
    They sound a bit crude but better by far than nothing.
    Getting "serious" (ie., desperate) with diagnostics you can have a
    look at:
    http://www.scyld.com/diag/index.html
    In a pinch, and much, much easier, can you borrow a different nic for
    testing? And change out the patch cord between your nic and CM?

    As I said in my earlier post, I have no info or experience with
    Toshiba cable modems. You can access CM setup/signal data via
    http://192.168.100.1 (?). What configuration and signal readings does
    its diagnostic page show? Borderline signals can be cause erratic
    behavior/errors. Have you checked for firmware updates? (While
    waiting I did look but found none offered -- they are expected to be
    delivered by ISP when CM boots, probably)

    Hardware problems would have to be pretty subtle, I think, to show in
    Linux but not Windows. Combined with different drivers, though, you
    can't rule them out. Also some Windows drivers/OS updates are
    suspected to leave some few nics with scrambled registers that cause
    weird symptoms or complete failure under Linux.

    traceroute returns (same results with -I) to name servers ...

    With $ /usr/sbin/traceroute 66.189.219.29 or .30
    I get as far as 12.127.79.150 (10th hop) which http://openrbl.org
    shows as:
     Lookup 12.127.79.150 (unresolved) in 21+11 Zones
      AS: [NO_ROUTE]
     Net 12/8 ATT Middletown, New Jersey @ip.att.net
    ie., packets are being dropped quick...

    With $ /usr/sbin/traceroute 66.169.254.29 or .30
    I get as far as:
    10 12.127.79.150 (12.127.79.150) 81.083 ms 79.638 ms 88.969 ms
    11 * * *
    12 * * *
    13 172.29.237.161 (172.29.237.161) 87.306 ms 105.076 ms 86.204 ms
    14 172.29.237.166 (172.29.237.166) 102.917 ms 101.913 ms 102.467
    ms
    15 192.168.3.2 (192.168.3.2) 92.433 ms 84.910 ms 104.389 ms

    http://www.openrbl.org shows:
     Lookup 172.29.237.161 (unresolved) in 21+11 Zones
      AS: 172.28.0.0/15 AS3549 Global Crossing Sunnyvale/California
     Net 172.16-172.31 IANA-BBLK-RESERVED Marina Del Rey, California
    and
     Lookup 172.29.237.166 (unresolved) in 21+11 Zones
      AS: 172.28.0.0/15 AS3549 Global Crossing Sunnyvale/California
     Net 172.16-172.31 IANA-BBLK-RESERVED Marina Del Rey, California

    and since you're in the area of:
    from Linux:
     Lookup 68.113.7.247 (ts46-01-qdr2036.wlawla.wa.charter.com)
            in 21+11 Zones
      AS: [NO_ROUTE]
     Net 68.113.0-31 KWCK-WA-68-113-0 St Louis, Missouri @chartercom.com
    and from Windows:
     Lookup 66.189.180.71 (ts46-01-qdr583.wlawla.wa.charter.com)
            in 21+11 Zones
      AS: [NO_ROUTE]
     Net 66.189.176-191 MDFRD-OR-66-189-176 St Louis, Missouri
    @chartercom.com

    Seems "strange" to me that you could end up in two such separated
    networks -- Kennewick and Medford -- at the same time from Walla
    Walla. Maybe they send Linux users to ISP's mud hole net ...

    Name server IPs run through openrbl.org:
     Lookup 66.189.219.29 (ns1.wa.charter.com) in 21+11 Zones
      AS: [NO_ROUTE]
     Net 66.189.208-223 PTORCD-WA-66-189-208
            St Louis, Missouri @chartercom.com

     Lookup 66.169.254.29 (unresolved) in 21+11 Zones
      AS: [NO_ROUTE]
     Net 66.169.224-255 KWCK-WA-66-169-224
            St Louis, Missouri @chartercom.com

    This is about as far as you can get with "easy" tools. The next step
    up is to pull out Ethereal and capture all packets as you ping and
    traceroute to your name servers (and GW if you want). Keep capturing
    as you use Moz and some other browser to visit a few "problem" sites.

    Normally you'll capture 500 - 1000KB per hour so there shouldn't be
    any worry with data size, should there? This is just "test" browsing
    -- don't get carried away.

    If you do have Konqueror available, compare it to Moz performance
    (while capturing packets). It uses different technology and if it or
    another browser works consistently but Moz does not, then Moz is
    boinking out on you. If eyeball performance is similar (and not good)
    it's time to inspect those packets you captured. Save them to file to
    keep around and we can give pointers on how to use Ethereal for
    revealing nuggets of information.

    My turn-around time on the NGs is deplorable -- life in rural America
    leaves me only with Google as a means of access. But I am watching...

    hth a little,
    prg
    email above disabled


  • Next message: David Efflandt: "Re: Expanding Wireless coverage by cascading APs"

    Relevant Pages

    • RE : Router problems on Redhat 9.0 Linux 2.4.20-13.9.HOSTAP
      ... General Red Hat Linux discussion list ... >> Your wlan range is inside that so no routing is performed. ... > The machines on the wireless network should have the default gateway ... it looks like the you can ping from a machine on ...
      (RedHat)
    • RE : Router problems on Redhat 9.0 Linux 2.4.20-13.9.HOSTAP
      ... General Red Hat Linux discussion list ... > Your wlan range is inside that so no routing is performed. ... it's impossible to ping a machine on ... 10.1.10.1, while the machines on LAN including the linux box/router, ...
      (RedHat)
    • Re: Fedora 4 Routing table question
      ... Fedora 4 Routing table question ... maybe it isn't my routing table. ... boxes can ping the Linux box and the Linux box can ping the windows ...
      (Fedora)
    • Re: ICS + Linux
      ... I have an internet connection on a windows ME box that is running ICS to an adapter which has a crossover cable directly to my linux box. ... The mini-network is set up properly (I can ping 192.168.0.1(my ME box) and it can ping 192.168.0.2). ... poly-p man ...
      (comp.os.linux.networking)
    • Re: Two interfaces on the same network
      ... Each machine can access the html access on the Linux box. ... > they will be on the same subnet for that scenario. ... Ping test: ... If eth0 is unplug, ...
      (RedHat)