Page 1 of 2 12 LastLast
Results 1 to 10 of 15

Thread: Gatling: High-Performance Open-Source Server

  1. #1
    Join Date
    Jan 2007
    Posts
    14,369

    Default Gatling: High-Performance Open-Source Server

    Phoronix: Gatling: High-Performance Open-Source Server

    After recently bringing up the G-WAN web-server, a Phoronix reader brought up Gatling. The Gatling software is a high-performance open-source web-server not known as widely as Apache, Nginx, or lighttpd...

    http://www.phoronix.com/vr.php?view=MTE0NDQ

  2. #2
    Join Date
    Jun 2006
    Posts
    351

    Default

    No ARM builds?

    Monkey got them, even running on the Raspberry Pi a setup I'll try as soon as I decide what fsking SD card I should purchase.

    So these news items mean that a lightweight web server head-to-head is maybe in the works? I would love to see one, even one on your ARM pandaboard cluster... yummy....

  3. #3
    Join Date
    Jun 2010
    Location
    ฿ 16LDJ6Hrd1oN3nCoFL7BypHSEYL84ca1JR
    Posts
    1,023

    Default

    I have used that sometimes and I like it for its simplicity. Want to serve a specific directory quicky?
    Code:
    cd directory; gatling -S -F -p 1234
    Want to serve content per cgi/fcgi? Couldn't be easier (but its not as clear documented as I'd like, you need to run it as root I believe, not a problem since it supports privilegue dropping).

    I think he wrote gatling mainly as a "real world" usage of his diet libc. (Some slides about the why).

    At the moment it seems it isn't working correctly anymore due to not having been updated for a while, probably.

    For only one concurrent request at a time it works fine with ~4k requests/second, but with more concurrent requests there are some problems with sockets:
    Code:
    socket_error 32 Bad file number
    close/writefail 32 98 218
    Also, nginx does more requests/second with 1 concurrent request.

    I only tested a small static file with the content "hello world".

    On the other hand his website and blog run on gatling very, very reliably so I would guess it's just a problem with new linux behaviour.

  4. #4
    Join Date
    Oct 2009
    Posts
    101

    Default

    OK, this is not particularly relevant to its performance, but CVS? Seriously? In 2012? For an open-source project? I have to use CVS at work, and it is...not my favorite part of my job. I have to wonder how many potential contributors will be turned off by CVS's limitations and quirks when I can barely tolerate it when I get paid to do so.

  5. #5
    Join Date
    Feb 2010
    Location
    Srpska Republic, Bosnia&Herzegovina
    Posts
    106

    Default

    A few things I've noticed:

    - "file not found" error if there's no trailing "/" on directory name
    - directory listing does not list directories first
    - I also get some socket errors

  6. #6
    Join Date
    Jun 2010
    Location
    ฿ 16LDJ6Hrd1oN3nCoFL7BypHSEYL84ca1JR
    Posts
    1,023

    Default

    Quote Originally Posted by waucka View Post
    I have to wonder how many potential contributors
    Well, first you have to know that it's by no means a project with much contribution anyway, only almost only a one-man project:
    Code:
     ~/gatling (cvs)-[gatling] % cvs log | grep author | grep leitner -c
    980
     ~/gatling (cvs)-[gatling] % cvs log | grep author | grep -v leitner -c
    5
    There are hardly any websites using it and from googling I learned that back in 2003 it was written to experiment with new APIs (Benchmarks from back then) and another goal was to write a webserver that would survive being slashdotted.

  7. #7
    Join Date
    Jul 2008
    Location
    Berlin, Germany
    Posts
    821

    Default

    Thanks to the fact that gatling does not attempt to support dynamic content, there needs not be any fast-paced development.

    The Monkey benchmarks are of limited value, as the exact build parameters for nginx are not known. Normally you build a monolithic binary with exactly the features you want.

  8. #8
    Join Date
    Jan 2009
    Posts
    462

    Default

    I'm really curious as to why this even necessitated an article. What does Gatling have that Apache httpd does not offer, or can not be achieved with simple compile time options? While a book can be written on the reverse perspective, I've always wondered what makes all of these lightweight httpd servers attractive.

    F

  9. #9
    Join Date
    Feb 2008
    Location
    Linuxland
    Posts
    4,994

    Default

    Quote Originally Posted by chithanh View Post
    The Monkey benchmarks are of limited value, as the exact build parameters for nginx are not known. Normally you build a monolithic binary with exactly the features you want.
    Defaults, IIRC. ./configure

  10. #10
    Join Date
    Jul 2008
    Location
    Berlin, Germany
    Posts
    821

    Default

    Quote Originally Posted by russofris View Post
    I'm really curious as to why this even necessitated an article. What does Gatling have that Apache httpd does not offer, or can not be achieved with simple compile time options? While a book can be written on the reverse perspective, I've always wondered what makes all of these lightweight httpd servers attractive.
    Gatling has a totally different use case from Apache. No amount of compile-time options will make Apache into a 125k static x86 binary. As it says on the homepage, gatling was developed to sustain the Slashdot effect and similar flash crowd events.

    Quote Originally Posted by curaga View Post
    Defaults, IIRC. ./configure
    So this means the compiler flags were empty, ie. without any optimization at all? Also configure scripts typically detect what is installed on the system and adjust the build based on that.

Posting Permissions

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