1 reply to this topic

[fatboy]

Anyone here developing with this yet? As an old RDBMS guy I'm trying to reconcile the two. I think it's easy enough and obvious how to do. Just wondering if anyone else here is playing with it.

P.S. I'm using Pig and Hive on top of it. Just 'cause it makes a lot of sense.

Edited by fatboy, 14 September 2009 - 11:01 PM.

[fatboy]

Missing the point:
MapReduce: A Major Step Backwards

Almost getting the point:
Databases are hammers: MapReduce is a screwdriver

In the good old days when relational theory was widely accepted but RDBMS were still immature (I'm talking mid-to-late 80's here, 15-20 years or so post-Codd and Oracle and Microsoft were just other startups I couldn't afford to invest in), we implemented our relational designs by using separate files for tables and indices and hard-coding our own pre-defined queries against them.

A project I am working on today involves a 50+ terabyte read-only dataspace with pre-defined queries and processing.

I've already started letting my hair grow long again and recently found my old bell-bottoms (although the hair and jeans might've been the 70's ... it's all a blur). What's old is new and it's deja vu all over again.

Edited by fatboy, 18 September 2009 - 09:43 PM.