2.5 Million Requests

Posted by chad March 07, 2006 @ 04:06 PM

In case you were wondering if Rails can scale: Eric Hodel reports that the Robot Co-op served 2,587,240 requests through their Rails applications last Saturday.

Do you have any scalability stories to share?

Posted in General | 30 comments

Comments

  1. Ed Powell on 07 Mar 17:52:

    What, all at once? :)

  2. Kyle on 07 Mar 18:38:

    Ed: Yep! They’re not as popular as they think… I just ran:

    ab -n 2587240 -c 2587240 http://43things.com/

    It’s amazing that my box was able to push that much traffic, but hey, its a macbook pro.

  3. dylan on 07 Mar 18:59:

    hahahaha… classic Kyle ! can’t wait to try that on my friends quad G5 !!

    Congrats Eric and Robot Co-op team !

  4. Jeet on 07 Mar 19:15:

    BTW What is an ab command??

  5. Bugsy on 07 Mar 19:35:

    ap – Apache server benchmark tool. http://httpd.apache.org/docs/2.0/programs/ab.html

  6. thomas.baustert@b-simple.de on 07 Mar 19:35:

    Great!

    Any information about the hardware behind available?

  7. Piotr Usewicz on 07 Mar 19:35:

    ab stands for Apache Benchmark ;] A little app which does dozens of requests to a webserver.

  8. Bugsy on 07 Mar 19:36:

    That’s ab. :)

  9. Piotr Usewicz on 07 Mar 19:36:

    ab stands for Apache Benchmark ;] A little app which does dozens of requests to a webserver.

  10. jay on 07 Mar 19:53:

    any idea how many servers they needed to handle that?

  11. ceejayoz on 07 Mar 20:45:

    Yeah, as jay says, those numbers aren’t telling us much without some indication of what they’re running on.

  12. Duane Johnson on 07 Mar 23:34:

    ceejayoz: The point is that Ruby on Rails can work for that many requests in a day, which really is telling us something. It’s not a question of “my language / framework / c program” can do it faster, because everyone knows Rails is not blazingly fast.

    But knowing that it’s fast enough is newsworthy, and good news indeed.

    Having said all that, I’d like to know how to duplicate that kind of throughput.

  13. Scott on 08 Mar 05:13:

    How many views does Penny Arcade get per hour || day?

  14. Scott on 08 Mar 05:13:

    How many views does Penny Arcade get per hour || day?

  15. Jaru on 08 Mar 11:20:

    Wonderful!!

    This amazing performance is done not only by Rails by DHH but also by Database Power. Rails is in very close relationship with Database. Hooray DHH and Hooray Database vendors !!

  16. Jaru on 08 Mar 11:20:

    Wonderful!!

    This amazing performance is done not only by Rails by DHH but also by Database Power. Rails is in very close relationship with Database. Hooray DHH and Hooray Database vendors !!

  17. Jaru on 08 Mar 11:21:

    Wonderful!!

    This amazing performance is done not only by Rails by DHH but also by Database Power. Rails is in very close relationship with Database. Hooray DHH and Hooray Database vendors !!

  18. JBrickley on 08 Mar 12:21:

    What we really need is a whitepaper detailing the best way to cluster both hardware and software in a full on scaled production enivronment. Then real world engineering measurements detailing real hits and not simulated hits. Not just cached pages and images but full database inserts and queries along with page renderings.

    - 10 1U dual processor rack servers (like XServes or Dell) - SAN setup for disk space. - 4 of them running PostgreSQL/Oracle setup in a cluster. - 3 running FCGI or the newer SCGI processes. - 3 running Lighttpd instances. - Apache front end.

    Lots of caching and tuning, etc. Real performance numbers from a big site.

    Unfortunately, all this seems to be black magic to most of us. I’ve seen reports of TextDrive having issues with Rails systems running Apache and FCGI – getting Slashdotted on one of their shared host rails sites and switching to Lighttpd then being amazed that it absorbed all the traffic and could have handled a whole lot more! They seem to still use Apache as a Proxy to backend Lighttpd instances.

  19. null on 08 Mar 13:19:

    This double post bug is very annoying. I wonder why it takes so long to fix it.

  20. byronsalty on 08 Mar 13:45:

    multiple posts are annoying…

    JBrickley: That sounds like quite a setup – what are you handling with all that?

  21. Matt on 08 Mar 15:24:

    From what I hear, double posts are a Firefox bug. Make sure you have the latest Firefox release.

  22. Jeremy on 08 Mar 15:27:

    http://www.penny-arcade.com

    I imagine they get a lot of traffic on their RoR site.

  23. null on 08 Mar 23:12:

    test

  24. Zachery Hostens on 09 Mar 07:55:

    JBrickley, why would you take 3 amazingly fast lighttpd box’s and throw them behind 1 apache instance… that would almost defeat the purpose, as well just wth is apache doing in this scheme?

    a better idea would have those 4 rails box’s running memcache for session(/ar persistance?) storage, the 3 lighttpd front-ends, and a dedicated hardware load-balancer.

    there is little to no reason to having a web server (apache) proxy to a web servers (lighttpd box’s) proxy to fcgi processes.

    just my 2 cents

  25. M on 09 Mar 09:29:

    Unless, Zachery, Apache has some modules that solves a problem for you.

  26. m on 11 Mar 00:45:

    offtopic: why on bloody earth havent they added validates_uniqueness_of to comments! this post is duplicatedspamed!

  27. giandrea on 11 Mar 02:07:

    m: probably because nobody could post “ok!” or “yes!” twice! O_O

  28. null on 11 Mar 12:02:

    offtopic: Same name and same comment? Duplicated. Different name and same comment. Ok. It’s not rocket science…

  29. Whatever on 14 Mar 10:37:

    Hey look, somebody made a generic blog site that does absolutely nothing of interest and it was able to keep up with a moderate number of requests. This is revolutionary.

  30. Maedi on 18 Mar 10:50:

    There are some angry people around the web aint there?