Map Reduce vs. Parallel DBMS

• A Comparison of Approaches to Large-Scale Data Analysis

Core points of the article

Loading data into a DBMS is slower:

Executing jobs on parallel DBMS is faster (at least 3 times on a 100 node cluster):

As we demonstrate in Section 4, at 100 nodes the two parallel DBMSs range from a factor of 3.1 to 6.5 faster than MapReduce on a variety of analytic tasks. While MR may indeed be capable of scaling up to 1000s of nodes, the superior efficiency of modern DBMSs alleviates the need to use such massive hardware on datasets in the range of 1–2PB.

eBay is used as an example for handling huge amounts of data in a parallel DBM cluster:

eBay’s Teradata configuration uses just 72 nodes (two quad-core CPUs, 32GB RAM, 104 300GB disks per node) to manage approximately 2.4PB of relational data.

Fault tolerance is much easier when doing MapReduce computations:

The MR frameworks provide a more sophisticated failure model than parallel DBMSs. While both classes of systems use some form of replication to deal with disk failures, MR is far more adept at handling node failures during the execution of a MR computation.

Discussion of schema vs. schema free:

One might think that the absence of a rigid schema automatically makes MR the preferable option. For example, SQL is often criticized for its requirement that the programmer must specify the “shape” of the data in a data definition facility. On the other hand, the MR programmer must often write a custom parser in order to derive the appropriate semantics for their input records, which is at least an equivalent amount of work. But there are also other potential problems with not using a schema for large data sets.can query this structure to discover such knowledge.

In general, the SQL DBMSs were significantly faster and required less code to implement each task, but took longer to tune and load the data.

The article is worth the read and holds many more points that I don't really want to copy-paste in. So if you found the quotes interesting, read the article ;-)