Writing MapReduce is still too complex and daunting to many programmers. So, not surprisingly, there are currently a lot of initiatives to simplify it and this will continue.
The following points state my opinion:
- With every new release, high-level languages such as Pig and Hive will be faster and will contain more libraries. For example, new release of Hive has a cost-based optimizer, which makes the query up to 2.5 times faster.
- MapReduce patterns will be more defined and there will be ready-made templates to choose from.
- There are tools already in the market for code-free MapReduce programs. It is likely that more organizations will start using these tools.
- Functional programming Scala and frameworks such as Cascading and Scalding are now more popular in the big data world. I would encourage readers to find out why.
- Machine learning, although I didn't cover it in this book, is a growing topic. With projects such as Apache Mahout and Spark MLlib, predictive modeling and machine learning will be used even more for clustering, recommendations, and fraud detection.
- Analytics tools such as R are on the path to be executed on native Hadoop. This will change the analytics landscape as the big data will be analyzed directly on Hadoop itself without moving it out of Hadoop.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.