The Community Forums

Interact with an entire community of cPanel & WHM users!
  1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

performance of razor, DCC, SARE and bayes

Discussion in 'General Discussion' started by abubin, Jul 1, 2005.

  1. abubin

    abubin Well-Known Member

    Joined:
    Dec 7, 2004
    Messages:
    393
    Likes Received:
    1
    Trophy Points:
    18
    I am now ready to install one of these services on my spamassassin but would like to know how effective and heavy load these would affect my server.

    razor
    DCC
    SARE
    bayes

    For eg. I have tried bayes before and find that with proper spam/ham learning, the spam detection is already so high that you don't really need to install razor, DCC or SARE. But then maybe with SARE, it's already good enough to fight spam that other "addons" for spamassassin are not needed.

    Please give your comment and your experience in using them.

    Thanks in advance.
     
  2. gupi

    gupi Well-Known Member

    Joined:
    Apr 27, 2004
    Messages:
    125
    Likes Received:
    0
    Trophy Points:
    16
    I am very satisfied about the results.
    Just sit and watch the logs.
    (followed RVSkin tutorial)

    Speed: good
    Accuracy: excellent
     
  3. abubin

    abubin Well-Known Member

    Joined:
    Dec 7, 2004
    Messages:
    393
    Likes Received:
    1
    Trophy Points:
    18
    speed of which one? You mean all of them combined?
     
  4. gupi

    gupi Well-Known Member

    Joined:
    Apr 27, 2004
    Messages:
    125
    Likes Received:
    0
    Trophy Points:
    16
    Yes, all combined.
    After following the RVSkin antispam tutorial, I had the luck of a massive spam delivery on one of hosted domains. Watching the logs I was very impressed about the filters, blocking almost everything. Of course, in the tutorial the SPAM level is set to 12; this level may vary in certain cases, but after fine tuning, you can get excellent results.
    However, I do not have a *huge* mail traffic on the server; an average of 1K e-mails/day, for all domains.
     
  5. abubin

    abubin Well-Known Member

    Joined:
    Dec 7, 2004
    Messages:
    393
    Likes Received:
    1
    Trophy Points:
    18
    hmm...1K mails per day for valid mails or includes all invalid mails?

    We have about 1K mails as well after RBL on MTA level (over 60k blocked by RBL). So, of these 1K mails, more than 60% are spams detected by spamassassin. Without all the extra implementation, i would say the accuraccy is about 80%. Would love to increase the accuraccy to above 90%.

    Would appreaciate if someone who had experience on these tools give some comment.

    One thing I am thinking is using RAZOR because it's spam detection is submitted by users. I am thinking this would be as accurate as bayes even though they uses different method of detection. Just want to know how much load RAZOR actaully take on the server. Also, any problems people are facing with these addons for spamassassin.
     
  6. gupi

    gupi Well-Known Member

    Joined:
    Apr 27, 2004
    Messages:
    125
    Likes Received:
    0
    Trophy Points:
    16
    I have this server for less than 1 month, and couldn't make very accurate statistics;
    anyway, looking at /.../exim_maillog scrolling for about 2 hours/day, one week, I can state that mail accuracy in my case is over 90%.

    I can state this since there are pretty few domains and I came to know the habits and 'valid' mail adresses.

    One more thing: I had this weekend one of my customers infected, sending about 1 message/second. Guess what: all mails were rejected (at HELO check, not at clamd), protecting the outer network and my server too. The customer was disconected/alerted, solved the issue, conected again.
     
Loading...

Share This Page