Why are column oriented databases so much faster than row oriented databases?
Be sure to read all the comments. Some of the techniques described are covered by patents (according to the comments) but there are open source implementations of alternatives. There is also a good discussion of the trade-offs in using this technique.
Search terms: Hybrid Word Aligned Bitmaps, HWAB, EWAB, FastBit.
Follow the links in the post and comments for more resources.
Questions:
From a topic map perspective, how would you structure a set of relational tables to represent the information items defined by the Topic Map Data Model? (Yes, it has been done before but no peeking! Your result will likely be very similar but I am interested in how you would structure the data. (If you want to think ahead, same question for the various NoSQL options.)
For the relational database, how would you structure a chain of selects to choose all the information items that should merge for any particular item. In other words, start off with the values of an item that should merge and construct a select that gathers up the other items with which it should merge.
Enumerate the operations you would need to perform post-select to present a “merged” information item to the final user.
Observation: Isn’t indexing the first step towards merging? That is we have to gather up all the relevant representatives of a subject before we can consider the mechanics of merging?
First seen at myNoSQL.