FTP transfer times. Linux vs. XP - Linux

This is a discussion on FTP transfer times. Linux vs. XP - Linux ; Now that the mouse pad on my laptop is behaving better I thought I'd try a few things. One thing that I did was to download a 250 Meg binary file (it's a compressed .tar.gz file) to the laptop using ...

+ Reply to Thread
Results 1 to 12 of 12

Thread: FTP transfer times. Linux vs. XP

  1. FTP transfer times. Linux vs. XP

    Now that the mouse pad on my laptop is behaving better I thought I'd try a
    few things. One thing that I did was to download a 250 Meg binary file
    (it's a compressed .tar.gz file) to the laptop using both Windows and
    Linux.

    The FTP server is a network attached storage drive that runs some sort of
    firmware (likely Linux) that supports both FTP and SMB protocols. I copied
    the 250 meg file to the FTP directory and timed the downloads. The times
    were manually recorded using a stopwatch (iPod touch) but they're pretty
    accurate.

    Windows explorer 3:25
    Windows CLI ftp 3:24 Linux CLI ftp 3:19

    The wireless access point is a Linksys and it's running DD-WRT firmware.
    There wasn't any other network traffic at the time.

    Someday when I'm feeling motivated I'll plug in the ethernet cable into
    the laptop and run the test again.


    --
    Ubuntu Linux
    17:52:42 up 17 days, 8:01, 1 user, load average: 0.06, 0.04, 0.05

    Don't you feel more like you do now than you did when you came in?


  2. Re: FTP transfer times. Linux vs. XP

    In comp.os.linux.advocacy, Ezekiel

    wrote
    on Tue, 14 Oct 2008 17:59:01 -0400
    :
    > Now that the mouse pad on my laptop is behaving better I thought I'd try a
    > few things. One thing that I did was to download a 250 Meg binary file
    > (it's a compressed .tar.gz file) to the laptop using both Windows and
    > Linux.
    >
    > The FTP server is a network attached storage drive that runs some sort of
    > firmware (likely Linux) that supports both FTP and SMB protocols. I copied
    > the 250 meg file to the FTP directory and timed the downloads. The times
    > were manually recorded using a stopwatch (iPod touch) but they're pretty
    > accurate.
    >
    > Windows explorer 3:25
    > Windows CLI ftp 3:24 Linux CLI ftp 3:19
    >
    > The wireless access point is a Linksys and it's running DD-WRT firmware.
    > There wasn't any other network traffic at the time.
    >
    > Someday when I'm feeling motivated I'll plug in the ethernet cable into
    > the laptop and run the test again.
    >


    Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s.
    This is well below Wireless-G of 54 Mbps suggested by

    http://www.linksys.com/servlet/Satel...VisitorWrapper

    though is about the right speed for Wireless-B (11 Mbps).

    --
    #191, ewill3@earthlink.net
    Murphy was an optimist.
    ** Posted from http://www.teranews.com **

  3. Re: FTP transfer times. Linux vs. XP

    Ezekiel wrote:

    > The FTP server is a network attached storage drive that runs some sort of
    > firmware (likely Linux) that supports both FTP and SMB protocols. I copied
    > the 250 meg file to the FTP directory and timed the downloads. The times
    > were manually recorded using a stopwatch (iPod touch) but they're pretty
    > accurate.
    >
    > Windows explorer 3:25
    > Windows CLI ftp 3:24
    > Linux CLI ftp 3:19



    Why would you expect the times to differ? You are limited by bandwidth.
    Don't you know that people use 150-MHz Pentiums as webservers? You
    moved about 1 MB/sec. Hard drives operate above 50 MB/sec nowadays, and
    so do CPUs and RAM and ethernet. I don't know why you would expect the
    OS to be rate-limiting for FTP.

    If you want to see OS-dependent rates, try writing a C program that
    writes and reads a bunch of files.

  4. Re: FTP transfer times. Linux vs. XP


    "The Ghost In The Machine" wrote in message
    news:kqifs5-ugv.ln1@sirius.tg00suus7038.net...
    > In comp.os.linux.advocacy, Ezekiel
    >
    > wrote
    > on Tue, 14 Oct 2008 17:59:01 -0400
    > :
    >> Now that the mouse pad on my laptop is behaving better I thought I'd try
    >> a
    >> few things. One thing that I did was to download a 250 Meg binary file
    >> (it's a compressed .tar.gz file) to the laptop using both Windows and
    >> Linux.
    >>
    >> The FTP server is a network attached storage drive that runs some sort of
    >> firmware (likely Linux) that supports both FTP and SMB protocols. I
    >> copied
    >> the 250 meg file to the FTP directory and timed the downloads. The times
    >> were manually recorded using a stopwatch (iPod touch) but they're pretty
    >> accurate.
    >>
    >> Windows explorer 3:25
    >> Windows CLI ftp 3:24 Linux CLI ftp 3:19
    >>
    >> The wireless access point is a Linksys and it's running DD-WRT firmware.
    >> There wasn't any other network traffic at the time.
    >>
    >> Someday when I'm feeling motivated I'll plug in the ethernet cable into
    >> the laptop and run the test again.
    >>

    >
    > Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s.
    > This is well below Wireless-G of 54 Mbps suggested by
    >
    > http://www.linksys.com/servlet/Satel...VisitorWrapper
    >


    The speed has always seemed slow to me. The system reports the laptop
    wireless to be "Intel(R) PRO/Wireless 2915ABG Network adapter" and both
    Linux and Windows both report that it's connected at 54 Mbps. But both have
    the same tranfer rates.


    > though is about the right speed for Wireless-B (11 Mbps).


    It does report a 54 Mbps connection. Signals strength is pegged since the
    WAP is about 20 feet away from me. It seems about right right "theoretical
    speed" for 802.11b but in practice you never get the full bandwidth out of
    any network. So here I'm getting an actual throughput of 10+ Mbps which is a
    little too high.

    The rest of the network is Gigabit ethernet with CAT-6 wiring and 1-Gig full
    duplex switches. Some devices like the WAP and only have 100 Mbps
    connections. Everything goes through the Gigbit switch and the 'connection
    speed' LEDs are all correct for the 1Gig vs 100Meg devices.



    > --
    > #191, ewill3@earthlink.net
    > Murphy was an optimist.
    > ** Posted from http://www.teranews.com **




  5. Re: FTP transfer times. Linux vs. XP


    "Matt" wrote in message
    news:s89Jk.72$Tu6.51@newsfe14.iad...
    > Ezekiel wrote:
    >
    >> The FTP server is a network attached storage drive that runs some sort of
    >> firmware (likely Linux) that supports both FTP and SMB protocols. I
    >> copied
    >> the 250 meg file to the FTP directory and timed the downloads. The times
    >> were manually recorded using a stopwatch (iPod touch) but they're pretty
    >> accurate.
    >>
    >> Windows explorer 3:25
    >> Windows CLI ftp 3:24
    >> Linux CLI ftp 3:19

    >
    >
    > Why would you expect the times to differ? You are limited by bandwidth.


    Not necessarily since I'm getting the data from my own home network. If I
    was getting better bandwidth through the 54 Mbps network connection then the
    efficiency of the driver and TCP/IP stack should be factor. If I ever do
    hardwired tests (the laptop is only 100Mbps) the difference between the
    Linux and Windows drivers and TCP stack should show some difference.


    > Don't you know that people use 150-MHz Pentiums as webservers?


    This is an FTP server on a dedicated NAS device. This isn't the bottleneck
    since the speed is about 20x faster if I FTP the file over a wired Gigabit
    connection.

    > You moved about 1 MB/sec. Hard drives operate above 50 MB/sec nowadays,
    > and so do CPUs and RAM and ethernet. I don't know why you would expect
    > the OS to be rate-limiting for FTP.


    See below.


    > If you want to see OS-dependent rates, try writing a C program that writes
    > and reads a bunch of files.


    Here's somewhat of a rhetorical question. Why would you expect the times to
    differ? In theory you should be limited by the bandwidth of the drives and
    controller. But in practice the times will be different because of the
    efficiency differences of the filesystem, drivers and OS. The same reason I
    wanted to measure FTP transfer times.




  6. Re: FTP transfer times. Linux vs. XP


    "The Ghost In The Machine" wrote in message
    news:kqifs5-ugv.ln1@sirius.tg00suus7038.net...
    > In comp.os.linux.advocacy, Ezekiel
    >
    > wrote
    > on Tue, 14 Oct 2008 17:59:01 -0400
    > :
    >> Now that the mouse pad on my laptop is behaving better I thought I'd try
    >> a
    >> few things. One thing that I did was to download a 250 Meg binary file
    >> (it's a compressed .tar.gz file) to the laptop using both Windows and
    >> Linux.
    >>
    >> The FTP server is a network attached storage drive that runs some sort of
    >> firmware (likely Linux) that supports both FTP and SMB protocols. I
    >> copied
    >> the 250 meg file to the FTP directory and timed the downloads. The times
    >> were manually recorded using a stopwatch (iPod touch) but they're pretty
    >> accurate.
    >>
    >> Windows explorer 3:25
    >> Windows CLI ftp 3:24 Linux CLI ftp 3:19
    >>
    >> The wireless access point is a Linksys and it's running DD-WRT firmware.
    >> There wasn't any other network traffic at the time.
    >>
    >> Someday when I'm feeling motivated I'll plug in the ethernet cable into
    >> the laptop and run the test again.
    >>

    >
    > Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s.
    > This is well below Wireless-G of 54 Mbps suggested by


    Oh... it's one of those "fun days" where I fixup a bunch of loose ends.
    Sluggish mouse, slow wireless and stuff like that. The only thing that I
    could think of was the WAP itself. The firmware that I had in there was:

    Firmware: DD-WRT v23 SP1 Final (05/16/06) std
    Time: 19:15:02 up 26 days, 15:09, load average: 0.07, 0.08, 0.04

    Hmmm... maybe there's updated frimware available. Sure enough there is so I
    download the latest stable and flash the WAP firmware to:

    Firmware: DD-WRT v24-sp1 (07/27/08) std
    Time: 19:18:56 up 1 min, load average: 0.42, 0.15, 0.05


    Oh yeah... the transfer times went from the 3:20 range to the 1:30.
    Certainly not blazing fast but I'll take a 2x improvement any day. Funny
    thing is that flashing the firmware seems to have reset the root password to
    the access point. Time to read the docs and see what they set it to.





  7. Re: FTP transfer times. Linux vs. XP

    On Tue, 14 Oct 2008 19:26:55 -0400, Ezekiel scribbled down:

    >
    > "The Ghost In The Machine" wrote in message
    > news:kqifs5-ugv.ln1@sirius.tg00suus7038.net...
    >> In comp.os.linux.advocacy, Ezekiel
    >>
    >> wrote
    >> on Tue, 14 Oct 2008 17:59:01 -0400
    >> :
    >>> Now that the mouse pad on my laptop is behaving better I thought I'd try
    >>> a
    >>> few things. One thing that I did was to download a 250 Meg binary file
    >>> (it's a compressed .tar.gz file) to the laptop using both Windows and
    >>> Linux.
    >>>
    >>> The FTP server is a network attached storage drive that runs some sort of
    >>> firmware (likely Linux) that supports both FTP and SMB protocols. I
    >>> copied
    >>> the 250 meg file to the FTP directory and timed the downloads. The times
    >>> were manually recorded using a stopwatch (iPod touch) but they're pretty
    >>> accurate.
    >>>
    >>> Windows explorer 3:25
    >>> Windows CLI ftp 3:24 Linux CLI ftp 3:19
    >>>
    >>> The wireless access point is a Linksys and it's running DD-WRT firmware.
    >>> There wasn't any other network traffic at the time.
    >>>
    >>> Someday when I'm feeling motivated I'll plug in the ethernet cable into
    >>> the laptop and run the test again.
    >>>

    >>
    >> Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s.
    >> This is well below Wireless-G of 54 Mbps suggested by

    >
    > Oh... it's one of those "fun days" where I fixup a bunch of loose ends.
    > Sluggish mouse, slow wireless and stuff like that. The only thing that I
    > could think of was the WAP itself. The firmware that I had in there was:
    >
    > Firmware: DD-WRT v23 SP1 Final (05/16/06) std
    > Time: 19:15:02 up 26 days, 15:09, load average: 0.07, 0.08, 0.04
    >
    > Hmmm... maybe there's updated frimware available. Sure enough there is so I
    > download the latest stable and flash the WAP firmware to:
    >
    > Firmware: DD-WRT v24-sp1 (07/27/08) std
    > Time: 19:18:56 up 1 min, load average: 0.42, 0.15, 0.05
    >
    >
    > Oh yeah... the transfer times went from the 3:20 range to the 1:30.
    > Certainly not blazing fast but I'll take a 2x improvement any day. Funny
    > thing is that flashing the firmware seems to have reset the root password to
    > the access point. Time to read the docs and see what they set it to.



    Transfer times with the updated access point firmware. One is from XP the
    other from Ubuntu. It doesn't matter which is which.


    ftp> get file.tar.gz
    local: file.tar.gz remote: file.tar.gz
    200 PORT command successful.
    150 Opening BINARY mode data connection for file.tar.gz(235971987 bytes).
    226 Transfer complete.
    235971987 bytes received in 83.12 secs (2772.3 kB/s)
    ftp> quit
    221 Bye-bye


    ftp> get file.tar.gz
    200 PORT command successful.
    150 Opening BINARY mode data connection for file.tar.gz(235971987 bytes).
    226 Transfer complete.
    ftp: 235971987 bytes received in 82.80Seconds 2850.01Kbytes/sec.
    ftp> quit
    221 Bye-bye



  8. Re: FTP transfer times. Linux vs. XP

    On Oct 14, 5:59*pm, Ezekiel wrote:
    > Now that the mouse pad on my laptop is behaving better I thought I'd try a
    > few things. One thing that I did was to download a 250 Meg binary file
    > (it's a compressed .tar.gz file) to the laptop using both Windows and
    > Linux.
    >
    > The FTP server is a network attached storage drive that runs some sort of
    > firmware (likely Linux) that supports both FTP and SMB protocols. I copied
    > the 250 meg file to the FTP directory and timed the downloads. The times
    > were manually recorded using a stopwatch (iPod touch) but they're pretty
    > accurate.
    >
    > * *Windows explorer 3:25
    > * *Windows CLI ftp *3:24 * Linux CLI ftp * *3:19
    >
    > The wireless access point is a Linksys and it's running DD-WRT firmware.
    > There wasn't any other network traffic at the time.
    >
    > Someday when I'm feeling motivated I'll plug in the ethernet cable into
    > the laptop and run the test again.
    >


    If you adjust all settings for pure speed (and code for no bottlenecks
    like writes to disk, etc), Linux wins hands down. The test is for pure
    maximum speed though, and not really applicable to most real world
    situations. Just in case anyone was wondering.

  9. Re: FTP transfer times. Linux vs. XP

    After takin' a swig o' grog, cc belched out
    this bit o' wisdom:

    > If you adjust all settings for pure speed (and code for no bottlenecks
    > like writes to disk, etc), Linux wins hands down. The test is for pure
    > maximum speed though, and not really applicable to most real world
    > situations. Just in case anyone was wondering.


    There've been other test under network congestion that indicate that
    Window is 65% the speed of Linux.

    But you have to take it all with a grain of salt, and remember that the
    speed you get may simply depend on how busy the server is.

    --
    Nature to all things fixed the limits fit,
    And wisely curbed proud man's pretending wit.
    As on the land while here the ocean gains,
    In other parts it leaves wide sandy plains;
    Thus in the soul while memory prevails,
    The solid power of understanding fails;
    Where beams of warm imagination play,
    The memory's soft figures melt away.
    -- Alexander Pope (on runtime bounds checking?)

  10. Re: FTP transfer times. Linux vs. XP

    On Tue, 14 Oct 2008 15:24:52 -0700, The Ghost In The Machine wrote:


    > Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s. This is
    > well below Wireless-G of 54 Mbps suggested by


    If I may intrude

    I've been making a living doing WIFI for the past 5 years, and have found
    that the maximum G speed obtainable is about 23 Mbits/s under the right
    circumstances, i.e usually a receive level of -60dBm minimum.

    This speed is measured using Iperf or Netperf on Linux boxes at each end
    of the wifi link.

    I simulate the signal attenuation using calibrated HP digital
    attenuators, and measure power levels on RF power measuring equipment.

    I personally test the thruput of all WIFI APs I sell, which is around 25
    a week.

    Speed drops off as signal level drops, with most consumer units having a
    cutoff around -85 dBm.


    >
    > http://www.linksys.com/servlet/Satel...ildpagename=US

    %2FLayout&cid=1133202177241&pagename=Linksys%2FCommon%2FVisitorWrapper
    >
    > though is about the right speed for Wireless-B (11 Mbps).



    Having a WiFi AP nearby is also no guarantee of a good signal, tho it
    usually means a stronger signal. The main problem is out of phase
    reflections which can cause terrible data rates. A indicator of this is a
    strong but unstable (varying) receive level.


    --
    Linux full time, on the desktop, since August 1997

  11. Re: FTP transfer times. Linux vs. XP


    "Terry Porter" wrote in message
    news:t9mdnen1qvsVHmHVnZ2dnUVZ_h6dnZ2d@netspace.net .au...
    > On Tue, 14 Oct 2008 15:24:52 -0700, The Ghost In The Machine wrote:
    >
    >
    >> Your figures above suggest a bandwidth of 1.22 MB/s - 1.25 MB/s. This is
    >> well below Wireless-G of 54 Mbps suggested by

    >
    > If I may intrude


    It's usenet.... intruding is par for the course. And in this case, you're
    helping quite a bit.



    > I've been making a living doing WIFI for the past 5 years, and have found
    > that the maximum G speed obtainable is about 23 Mbits/s under the right
    > circumstances, i.e usually a receive level of -60dBm minimum.


    The latest tests that I ran showed a tranfer rate of 22.2 Mbits/sec which is
    very close to the rate you're finding as the max obtainable. Faster would be
    nicer but I can certainly live with this.

    It's interesting that updating the DD-WRT firmware on my router is what
    fixed this problem. After upgrading the firmware (see my previous post) I
    suddenly saw a 2.5x increase in throughput with both Linux and Windows.


    > This speed is measured using Iperf or Netperf on Linux boxes at each end
    > of the wifi link.
    >
    > I simulate the signal attenuation using calibrated HP digital
    > attenuators, and measure power levels on RF power measuring equipment.
    >
    > I personally test the thruput of all WIFI APs I sell, which is around 25
    > a week.


    So out of curiousity... is there any brand that consistently performs well
    and any brand that you would avoid? All of my stuff if Linksys except for a
    Netgear gigabit switch.


    > Speed drops off as signal level drops, with most consumer units having a
    > cutoff around -85 dBm.
    >
    >
    >>
    >> http://www.linksys.com/servlet/Satel...ildpagename=US

    > %2FLayout&cid=1133202177241&pagename=Linksys%2FCommon%2FVisitorWrapper
    >>
    >> though is about the right speed for Wireless-B (11 Mbps).

    >
    >
    > Having a WiFi AP nearby is also no guarantee of a good signal, tho it
    > usually means a stronger signal. The main problem is out of phase
    > reflections which can cause terrible data rates. A indicator of this is a
    > strong but unstable (varying) receive level.


    I may be getting some reflections but my data rate seems pretty good right
    now. The WAP is in the basement mounted vertically to a sheet of plywood on
    the concrete wall. I do most of my wireless work in the room directly above.


    > --
    > Linux full time, on the desktop, since August 1997




  12. Re: FTP transfer times. Linux vs. XP

    Terry Porter wrote:

    > I personally test the thruput of all WIFI APs I sell, which is around 25
    > a week.


    You test every unit? Is there that much variability that you have to
    test every unit, not just one or two samples of each model?

    -- Mike --

+ Reply to Thread