MySQL at Facebook
is on Facebook.
To connect with MySQL at Facebook, sign up for Facebook today.
Sign UpLog In
Like
Profile Picture

MySQL at Facebook
106,303 likes · 958 talking about this

Highlights
  3. Recent Posts by Others on MySQL at FacebookSee All
    • Kushal Yadav
      I m a beginner to MySQL. I hav installed mysql-5.5.25-win32 and works fine in local host. but don't know how to connect to My web server db through nternet
      11 hours ago
    • Mark Callaghan
       Henrik has been been active recently. This was my favorite post. http://openlife.cc/blogs/2012/august/state-mysql-forks-conclusions
      71 · Saturday at 11:59am
  4. MySQL at Facebook shared a link.
    August 23
    TokuDB + MongoDB look good together. I like to learn how databases perform when data is much larger than RAM.

    http://www.tokutek.com/2012/08/10x-insertion-performance-increase-for-mongodb-with-fractal-tree-indexes/
    10x Insertion Performance for MongoDB with Fractal Trees | Tokutek
    www.tokutek.com
    10x Insertion Performance Increase for MongoDB with Fractal Tree Indexes. Large gains for NoSQL
    24 · Comment ·
    • Johannes Fly, Raduta Iulian and 132 others like this.
      • Mark Callaghan
        ‎15000 inserts/second will be a huge problem for Oracle and just about every other DBMS when the database gets larger than RAM and there are secondary indexes to maintain. Read the post linked here and setup a similar table with secondary i...
        ndexes. I expect the insert rate to fall off of a cliff when the database becomes larger than RAM. AFAIK, Oracle doesn't have anything like fractal indexes (as does TokuDB) nor does it have anything like the InnoDB insert buffer. Both of these features make inserts faster than disk reads for secondary index maintenance make things slow.

        But I am not sure that you would be allowed to publish the perf results. Does Oracle allow for benchmark results to be published?
        See More
        Friday at 9:21am
      • C Serhan Tezcan I don't think facebook engineers will be happy if they share the performance results of the system.
        Saturday at 4:44am · 1
  5. MySQL at Facebook shared a link.
    August 22
    I updated innosim. It now writes page numbers and checksums on pages and confirms the values when reading from disk. There are also options to simulate the overhead from page compression (see --compress-level). These changes make it better at simulating the IO behavior of InnoDB when evaluating new storage systems.

    http://bazaar.launchpad.net/~mdcallag/mysql-patch/mytools/files/head:/bench/innosim/
    http://bazaar.launchpad.net/~mdcallag/mysql-patch/mytools/files/head:/bench/innosim/
    bazaar.launchpad.net
    5 · Comment ·
  6. MySQL at Facebook
    August 22
    I used the Performance Schema in MySQL 5.6 to determine the source of mutex contention. I am very happy with the data it provided.

    The manual for the PS needs a short user's guide. The current content is more of a reference manual and the feature is complex.

    Mutex wait stats cannot be enabled dynamically and when enabled this reduced peak throughput by 5% to 15% during benchmarks. The overhead is too high to always be on, but we aren't willing to restart servers to enable collection when it is time to debug. I imagine many other people feel the same way about restart. Can this be changed?
    2 · Comment ·
  7. MySQL at Facebook
    August 16
    He works on VoltDB but his posts are usually about performance problems that are relevant to anyone working on high-performance servers.

    http://www.afewmoreamps.com/2012/08/great-lkml-thread-from-2k3.html
    1 · Comment ·
  9. MySQL at Facebook
    August 11
    While I am more interested in a MySQL storage engine for LevelDB, it is great to see new projects like these from MariaDB.

    https://www.facebook.com/MariaDB.dbms

    http://kb.askmonty.org/en/cassandra-storage-engine
    http://kb.askmonty.org/en/hbase-storage-engine
    MariaDB
    MariaDB is based on the popular opensource database MySQL, and it is a community developed, feature enhanced and backward compatible database solution. Find us at http://twitter.com/mariadb as well! We're also on Google+: https://plus.google.com/102059736934609902389/posts
    Page: 1,867 like this
    11 · Comment ·

Earlier in August

Show August

Earlier in 2012

Show 2012