GET 24/7 LIVE HELP NOW

Announcement

Announcement Module
Collapse
No announcement yet.

Oracle vs MySQL benchmark test

Page Title Module
Move Remove Collapse
X
Conversation Detail Module
Collapse
  • Filter
  • Time
  • Show
Clear All
new posts

  • Oracle vs MySQL benchmark test

    Hello,

    With reference to Oracle benchmark test detailed below, we've done a relatively simpler test between MySQL (on InnoDB) and Oracle 11g (Enterprise Edition) using similar H/W configuration but with a simpler DB schema which consists of only a few tables and no additional alternate index. The test program was runs (in multiple threads) to insert and update records onto the DB continuously until a total of 100M records were created and then the total time taken recorded in msec. As a result, we found that Oracle was able to deliver a performance of 86K insertion/sec whilst MySQL was delivering 28K insertion/sec (which is only about 1/3 of Oracle figure) both under the same H/W and SAN storage configurations.

    Wonder if anyone has any comments to this result and if this relatively big gap in performance a reasonable figure between MySQL and Oracle or would there be other suggestions for us to better tune the MySQL to make it performs a bit more better. Note that for test run with Oracle, we found that disk utilization reached upto more than 90% but only 60-70% for MySQL test.

    Thanks very much in advance.



    http://www.oracle.com/us/corporate/press/018144

    Oracle Database 11g Standard Edition One and HP ProLiant Deliver Highest Performance per Processor of any TPC-C Result ever Published
    Redwood Shores, CA – March 30, 2009

    News Facts
    Today, Oracle announced a new world record TPC-C benchmark for a two-socket system with Oracle® Database 11g Standard Edition One running on Oracle Enterprise Linux, (1) demonstrating the superior enterprise-class performance, scalability and flexibility of Oracle Database for customers of all sizes.

    Achieving 631,766 transactions per minute with a price/performance of $1.08/tpmC, Oracle Database 11g Standard Edition One running Oracle Enterprise Linux on an HP ProLiant DL370 G6 server equipped with two Intel® Xeon® x5570 quad-core 2.93 GHz processors, delivered the fastest result for a two-socket system as well as the highest transaction-per-minute-per-processor of any TPC-C result ever published. The storage subsystem consisted of HP StorageWorks MSA2324fc and MSA70 enclosures.

  • #2
    Dont know about Oracle but MySQL surely can do a lot better than 28K that you mention. Postgresql in 8.3 can do many multiple times over that number. Both are *much* lower in cost than an Oracle environment.

    Comment


    • #3
      Hello pkiula,

      Thanks for the comment.

      Would there be any further advices in relation to the fine tuning of the MySQL environment/configuration such that a much better result as you mentioned can be delivered by MySQL. The MySQL version we used in the benchmark test was 5.1.30 and both DB tests were run under a RHEL 5.2 (64 bits) environment.

      Comment


      • #4
        Awesome test.

        If you're using the out of the box config for innodb I would suggest the following changes ( keep in mind innodb_flush_log_at_trx_commit =2 will potenitally make you non-ACID compliant but faster, otherwise you're sync-ing the log buffer to disk with every commmit == bad), all of these settings I think will help your performance:

        transaction-isolation = READ-COMMITTED
        innodb_file_per_table
        innodb_flush_log_at_trx_commit = 2
        innodb_log_buffer_size = 8M
        # You can set .._buffer_pool_size up to 50 - 80 %
        # of RAM but beware of setting memory usage too high
        # Make this as big as you can up to 80%
        innodb_buffer_pool_size=256M
        innodb_flush_method=O_DIRECT

        Comment

        Working...
        X