Page 44 - Building Big Data Applications
P. 44

38 Building Big Data Applications

































                                         FIGURE 2.10 YARN execution flow.


               SQL/MapReduce interfacedthis is the layer that will create and distribute the jobs
                at the lowest MapReduce execution layer for
               SQL/MapReduce Librariesdcatalog of library functions

                The overall benefits of combining SQL/MapReduce include the following (Fig. 2.11):
               Use of SQL for powerful postresult analytics and MapReduce to perform large-scale
                data processing on unstructured and semistructured data
               Effectively use the sharding capabilities of MapReduce to scale up and scale out
                the data irrespective of volume or variety
               Provide the business user all the data with the same interface tool that runs on
                SQL

                The downside of the technology in evolution includes the following:
               Heavy dependency on custom libraries
               Current support on certain analytic functions
                The next generation of SQL/Mapreduce interfaces and libraries will solve a number of
             evolutionary challenges.
                The combination of HDFS and MapReduce creates an extreme architecture. What is
             important to note here is as follows:
   39   40   41   42   43   44   45   46   47   48   49