We all know that good, clean water is core to life. Without it, we can only survive for around three days. So what happens if your water source is polluted? Well, unless you filter the water sufficiently, there will definitely be some negative consequences. To get better results, you could enrich the water with fluoride, filter out the arsenic, and deliver it at the right pressure and temperature.

In a similar way, “unpolluted” data is core to a successful business – particularly one that relies on analytics to survive. But preparing data for analytics brings with it different requirements than storing data in a warehouse. How difficult is it to manage unfiltered data and get it ready for analytics?

The power of Hadoop is that it utilises schema on read. With a data warehouse, you often have to know what the tables look like before loading data. With Hadoop, you can pull data from any source or type and then figure out how to organize it. Organisations are beginning to use Hadoop as a dumping ground for all kinds of data because it is inexpensive and doesn’t require a schema on write. Such storage is often referred to as a Hadoop “data lake.” On the flip side, the Hadoop/MapReduce engine is not optimised for the iterative processing that analytics often requires. It is best suited to batch processing.