Find the answer to your Linux question:
Results 1 to 3 of 3
Morning, y'all, Long time reader, first time poster; got a question for ya: Our lab at school has been having some connection speed issues with downloads tunneled over SSH (SFTP, ...
Enjoy an ad free experience by logging in. Not a member yet? Register.
  1. #1
    Just Joined!
    Join Date
    Apr 2012
    Posts
    3

    Dot in username causes SSH speed issues


    Morning, y'all,

    Long time reader, first time poster; got a question for ya:

    Our lab at school has been having some connection speed issues with downloads tunneled over SSH (SFTP, SCP, etc.). I could get 30MB/s when I was downloading, but all my labmates could only get 5MB/s to the same site.

    After some fiddling, hunting, and testing, it turns out the problem was that their usernames had dots in them (according to our school's student naming convention) whereas mine didn't ( according to an older convention that I was named under). We are all using Fedora 15, which didn't give a peep when we created the user accounts using "system-config-user".

    My question is: why? I've found through some brief searching that some distributions don't like it when you create a user name with a dot, but that there are workarounds. I couldn't find any mention of it causing problems once the user accounts are created, and we've been running this way for about a year with no other issues.

    Any thoughts from you experts out there? 30MB/s without dot in username, 5MB/s with dots in username, everything else the same and no other issues.

    Thanks,
    RY

  2. #2
    Just Joined!
    Join Date
    Apr 2012
    Posts
    3
    More info:

    The problem only occurs inside of a terminal in the graphical shell (Gnome 3 in this case). Inside of an init 3, the issue does not present itself at all.

    However, if a user with a dot in a graphical session changes users (via su) inside of a terminal, the issue does not present itself either.

    Thoughts?

    Thanks,
    RY

  3. #3
    Just Joined!
    Join Date
    Apr 2012
    Posts
    3
    Well...looks like I figured it out. Looks like I didn't test as thoroughly as I thought I did. Wasn't obvious until we started fresh with a clean system. Turns out, all my labmates had used a default configuration option in $HOME/.ssh/config:
    TcpRcvBufPoll yes
    TcpRcvBuf 1000
    that shot our performance all to heck. I had not used the default configuration on any of my accounts, so I never saw that problem.

    Instead, use of:
    HPNDisabled no
    HPNBufferSize 2000
    is giving wonderful performance (from 3MB/s to >30MB/s)

    Hope this helps someone,
    RY

Posting Permissions

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