158 Advances in Mobile Cloud Computing Systems
TeraSort application sorts a large number of 100-byte randomly
generated records, and as such it consumes a considerable amount of
Hadoop resources in the form of computation, networking, and stor-
age I/O access. We first generate one million records, and store them
in the HDFS-backed storage with default replication factor three.
Since we have three DataNodes in the HDFS storage (one DataNode
in each zone), it is ensured that each zone has a copy of one million
records in its designated DataNode. In this setting, we run TeraSort
application on the stored records.
In Figure 3.13a, we visualize the progress of TeraSort job execution
with various map and reduce task assignment. The y-axisofthefigure
plots the ...