This chapter highlights the relational (row store) model for comparison and contrast with the NoSQL patterns to come. I think that the OLAP discussion and the role of NoSQL systems in that architecture is particularly valuable. I have not seen this discussed in other references.
In Hadoop, joins are very difficult to implement in Java MapReduce, so the data tends to be unnormalized. Using high level tools like Hive and PIg eases the join problem. However, once implemented, joins are highly scalable across commodity clusters. I wonder if, across actual deployments, it is a general characteristic that data in NoSQL systems tends to be more denormalized than in relational systems.
While the material on directory services and revision control systems was interesting, it seemed out of place in this chapter, and perhaps would better fit in one of the earlier technology introduction chapters.
In Hadoop, joins are very difficult to implement in Java MapReduce, so the data tends to be unnormalized. Using high level tools like Hive and PIg eases the join problem. However, once implemented, joins are highly scalable across commodity clusters. I wonder if, across actual deployments, it is a general characteristic that data in NoSQL systems tends to be more denormalized than in relational systems.
While the material on directory services and revision control systems was interesting, it seemed out of place in this chapter, and perhaps would better fit in one of the earlier technology introduction chapters.