Results 1 to 10 of 10

Thread: LST V3 - figures produced are not right nor consistent

  1. #1
    Newb
    Join Date
    Oct 2012
    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts

    LST V3 - figures produced are not right nor consistent

    I've bought LSTv2 and used it a lot, so when I saw V3 was out and it actually directly solved a problem I have (Ability to run properly from a command line against multiple fileservers) I bought it straight away. Well worth the pint of beer it costs.

    I use it to monitor the performance of a remote fileserver over a WAN link. It's been very useful as we've tweaked our fileserver config and seen a direct performance improvement (Tcp window size on netapp, but I digress)

    So I'm running a performance test from the command line, with teh settings:

    I'm getting for read 25.5 Mbps <- Which is pretty much the same as old versions and makes sense

    For Reading 1,9568 Mbps <- Which is impossible given the local LAN is only 100Mb and previous I would get ~80Mbps

    Here's the command line I used

    C:\Program Files (x86)\LAN Speed Test>LAN_SpeedTest.exe /D1 ("\\fileserver\fileshare$",0,10000,1000,0,1,0,1 )

    Or if I try another one

    C:\Program Files (x86)\LAN Speed Test>LAN_SpeedTest.exe /D1 ("\\anotherfileserver\data2$",0,1000000,10,0,1,0,1 )

    I get
    2017 Mbps for write and 100Mpps for read.

    I've done a bit more testing and, in summary, I think the figures reported are nonsense.

    The older tool provides reasonable, repeatable stats.

    What's up?

  2. #2
    Home Network Guru
    Join Date
    Dec 2011
    Posts
    120
    Thanks
    0
    Thanked 1 Time in 1 Post
    Hi Chaplic,

    Thanks for the comments and reported issue. You are the fourth person that has come across this now. I have been working with the others trying to figure out the scenario that was causing this. I was finally able to re-produce this here just now, and hope to have it fixed later tonight or tomorrow.


    -Pete

  3. #3
    Newb
    Join Date
    Oct 2012
    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Quote Originally Posted by Pete Totushek View Post
    Hi Chaplic,

    Thanks for the comments and reported issue. You are the fourth person that has come across this now. I have been working with the others trying to figure out the scenario that was causing this. I was finally able to re-produce this here just now, and hope to have it fixed later tonight or tomorrow.


    -Pete
    Good stuff, let me know if you want me to try anything

  4. #4
    Home Network Guru
    Join Date
    Dec 2011
    Posts
    120
    Thanks
    0
    Thanked 1 Time in 1 Post
    Good news! It's fixed Just go to http://totusoft.com/downloads.html and download the latest from there.

    3.1 (10/24/12) - Thirteenth Public Release
    -----
    B - In certain situations (Win7 to Win7 shared folder), the timer would stop after the packet was sent to the local cache instead of the drive's cache. This would cause the write results and throughputs to show a lot faster than it should.
    B - In certain situations (Win7 to Win7 shared folder), the last packet being read was already cached and would cause the read results and max throughput to show faster than it should.

    -Pete

  5. #5
    Newb
    Join Date
    Oct 2012
    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Quote Originally Posted by Pete Totushek View Post
    Good news! It's fixed Just go to http://totusoft.com/downloads.html and download the latest from there.

    3.1 (10/24/12) - Thirteenth Public Release
    -----
    B - In certain situations (Win7 to Win7 shared folder), the timer would stop after the packet was sent to the local cache instead of the drive's cache. This would cause the write results and throughputs to show a lot faster than it should.
    B - In certain situations (Win7 to Win7 shared folder), the last packet being read was already cached and would cause the read results and max throughput to show faster than it should.

    -Pete

    Thanks for this. I've briefly tested it, and it does appear to be more realistic. Great stuff. Thanks.

    Can you explain what exactly you mean by 'packet size', as this does have a drastic effect on the reported performance

    For example, with v1.1.7 writing a 20MB file to a remote fileserver I get reported write of 12Mbps and read of 50Mbps (which based on the intrastructure sounds reasonable)

    If on V3.1 I do:

    "C:\Program Files (x86)\LAN Speed Test\LAN_SpeedTest.exe"/D1 ("\\fileserver\share$",0,65536,200,0,1,1,1 )

    I get 6 and 18 for read and write


    Or

    "C:\Program Files (x86)\LAN Speed Test\LAN_SpeedTest.exe"/D1 ("\\server\share$",0,1000000,200,0,1,1,1 )

    I get close to the early version numbers (12.3 and 78.6) - again it's a live network and single test so believable

    But what reference does packet size have in relavance to TCP window size SMB packet size?

    Thanks
    Last edited by chaplic; 10-31-2012 at 05:39 AM.

  6. #6
    Home Network Guru
    Join Date
    Dec 2011
    Posts
    120
    Thanks
    0
    Thanked 1 Time in 1 Post
    Hi Chaplic,

    Packet size does not have anything to do with TCP Window Size. Packet Size in LAN Speed Test is really the file size in your case. In your first example, you would be testing with 200 packets (files) with the size of 65 KB each (it's actually 200 65 KB files packed into 1 file so it doesn't get your directory messy while testing). So, when testing with 200 - 65 KB files, the results are going to be considerably less than testing with 200 - 1 MB files as reading and writing larger files is much more efficient than reading and writing smaller files.

    -Pete
    Last edited by Pete Totushek; 11-01-2012 at 06:15 AM.

  7. #7
    Newb
    Join Date
    Oct 2012
    Posts
    4
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Quote Originally Posted by Pete Totushek View Post
    Hi Chaplic,

    Packet size does not have anything to do with TCP Window Size. Packet Size in LAN Speed Test is really the file size in your case. In your first example, you would be testing with 200 packets (files) with the size of 65 KB each (it's actually 200 65 KB files packed into 1 file so it doesn't get your directory messy while testing). So, when testing with 200 - 65 KB files, the results are going to be considerably less than testing with 200 - 1 MB files as reading and writing larger files is much more efficient than reading and writing smaller files.

    -Pete
    Ah, gotcha.

    So if I wanted to emulate the earlier version default I'd use

    0,20971520,1,0,1,1,1

    Test with one file of 20MB?

  8. #8
    Home Network Guru
    Join Date
    Dec 2011
    Posts
    120
    Thanks
    0
    Thanked 1 Time in 1 Post
    Exactly And if you'd like to log the results you would need to add one more ,1 at the end.

    -Pete

  9. #9
    Newb
    Join Date
    Jan 2013
    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Unhappy Same results in V3.4 I'm afraid

    I am getting these same results using LST V3.4. In a room of 14 identical Windows XP Dell Optiplex 390 PCs connected to the same switch I have run the speed test with varying results. Some of the pcs come back with 300+ Mbps reads that are what I expect then some report 7,000 Mbps! Strangely the 7,000 Mbps PCs are on a stopwatch test running slower. Any thoughts please?

  10. #10
    Home Network Guru
    Join Date
    Dec 2011
    Posts
    120
    Thanks
    0
    Thanked 1 Time in 1 Post
    Hi,

    What is your packet size and how many packets? Also, are you testing to LST Server or to a shared folder? If to a shared folder, are you sure that you have the right credentials in the folder name (ie. not testing to a local drive)? What happens when you bump your packet size up to 10MB? If you post some screenshots, that would help a lot. Or, email me at support@totusoft.com with more details.

    -Pete

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •