Slashdot Mirror
Oracle Claims Dramatic MySQL Performance Improvements
New submitter simula67 writes "Oracle wins back some karma from the open source community by releasing MySQL cluster 7.2 with ambitious claims of 70x performance gains. The new release is GPL and claims to have processed over 1 billion queries per minute. Readers may remember the story about Oracle adding commercial extensions to MySQL."
The case when the data set is bigger than RAM amount has not been investigated (link here, see the comments). The hard drive I/O speed would slow it dramatically, unless it's an expensive array of SSDs.
The latest (past year's worth) of PostgreSQL releases automatically rewrite subqueries as joins for you.
However, MSSQL 2008 performance dropped significantly from prior versions due to the exact opposite --- subqueries are now looped through, where previously they were handled with hash tables internally.
Slightly off-topic, but I recently had the oppurtunity to test the speed of a MySQL in-memory database. I have some frequently queried read-only data that simply would not handle the load in MS SQL and was looking for an in-memory solution. MySQL provided the simplest implementation - simply tell the table to use memory storage and configure the server to allow the amount of data you want to host (~250MB in this case). You also have to remember to reload the data from normal InnoDB tables every time you restart the server. I used the same table structures, keys indexes and stored procedures (almost the same) to query the data and linked it through MS SQL so that my applications never new the difference. On exactly the same hardware the speed increase was at least 50X over MS SQL.
Freedom of speech doesn't come with bandwidth.
No, rather it's amazing what happens when you architect the DB such that the entire thing is forcibly held in memory.