October 6, 2009
Oracle’s version of “actually, we’ve been doing MapReduce all along too”
In a recent blog post, Jean-Pierre Dijcks of Oracle makes the argument that Oracle has supported MapReduce all along, essentially because:
- You can do lots of procedural logic in the Oracle database, in a broad choice of languages, so in particular you can do Map steps.
- You can do lots of procedural logic in the Oracle database, in a broad choice of languages, so in particular you can do Reduce steps.
- Oracle offers a mechanism for parallelizing procedural logic.
Oracle doesn’t appear to have an explicit Map/Reduce programming interface, but I wouldn’t be surprised if Oracle Consulting cranked one out at some point to meet customer demand.
The post goes on to claim the usual in-database MapReduce benefit of avoiding the overhead of intermediate query result materialization. Presumably, then, Oracle’s quasi-MapReduce would also lack query fault-tolerance.
Categories: Analytic technologies, MapReduce, Oracle, Parallelization
Subscribe to our complete feed!
Comments
One Response to “Oracle’s version of “actually, we’ve been doing MapReduce all along too””
Leave a Reply
Did they do anything for you at the ociffe?Nope, not a multiple of 5!Funny story. In 1998 the division I am in decided to start doing 1, 5, 10, 15, 20, … year recognition awards.Someone had to write the application to generate the list of people each month that were to get the award.It was decided that it would just “start” – that is, if you were 5 years and one month – you would not get the new award, if you were 5 years – you would. If you were 4 years 11 months you would get it next month.I wrote that application September 1998.You guessed it, it was four years and eleven months before I got my recognition from that system :)Always thought it was funny that I wrote it, but missed the cutoff date by a month.