Data Normalization in moderation

by Sam T
In this article, the author, who also completed their masters project on data normalization, argues that although data normalization has it’s many benefits, too much of it can actually hinder the database instead of helping it. He brings up the main benefits of data normalization and shows the downsides to each benefit. In one example, normalization is used to help save space and cut down on data redundancy, the author argues that there’s really no point as hard drives are constantly being improved and as bigger hard drives are getting cheaper, normalizing some fields of a table wouldn’t help save space. In another example, normalization can help simply updates and reduce anomalies but he brings up the point that databases are constantly being read instead of being written. “If you have a table with relatively stable data that changes infrequently, normalization isn’t buying you a lot.” (Selikoff, 2008) He claims that data normalization is very useful, good database developers should find a good balance between using too much normalization and to see what the database needs.

This article related to our lecture this week because we discussed about the benefits of data normalization and touched on a couple of examples. Even though I was lost on the examples we did in class and this was the first time I’ve seen data normalization, I was a bit hesitant with agreeing with the author. I can understand why the author used the examples he used but I found a couple of flaws in his arguments.

In the first example I posted up above, he claims that space is constantly getting cheaper but without data normalization, space would constantly run out quicker. I can see his point that not every table needs data normalization but if every table didn’t have normalization, databases would quickly run out of space. This would require companies to keep spending resources and funding to purchase and install new drives. In the second example, he argues that some data rarely ever have to change, such as “United States of America” text in users table. Although the United States haven’t changed the name in about 200+ years there are many countries that have. Such as the Soviet Union to Russia or Yugoslavia to Serbia and Montenegro to finally Serbia. It would be easier to change a single country record than it is to change 10000+ user records. In the end of the article the author does admit that data normalization is very useful but only in moderation.

Selikoff, S. (2008, November 19). [Web log message]. Retrieved from http://www.selikoff.net/2008/11/19/why-too-much-database-normalization-can-be-a-bad-thing/

2 thoughts on “Data Normalization in moderation

  • October 20, 2012 at 8:00 pm
    Permalink

    As a beginner in database normalization and database in general, I would say that normalization is a pretty good thing. Yes, there can definitely be too much of a good thing, but I think in a majority of databases, normalization will benefit the database. It seems like the larger the database, the less you would need. Like in a world wide database of country names, the USA would be a country that would stay the same while other countries would take advantage of the normalization. But again, I am a novice.

  • October 20, 2012 at 8:39 pm
    Permalink

    Similar to the view in the article that I posted, normalization seems to be beneficial in tables that often get updated, however, for tables that are mostly static and the values hardly change, normalization might not be the best answer. In fact, cases like such and cases similar to what your article mentioned, the benefits of denormalization might outweigh the benefits of normalization.

Comments are closed.