Crazy, Wild Hadoop Prioritization Advice

November 12, 2015

I read “Top 10 Priorities for a Successful Hadoop Implementation.” A listicle. I understand. Clicks. Visibility. Fame. Fortune. Well, hopefully.

I wanted to highlight two pieces of advice delivered in a somber, parental manner. Here are two highlights from the write up intended to help a Hadoop administrator get ‘er done and keep the paychecks rolling in.

Item 2 of 10: “Innovate with Big Data on enterprise Hadoop.” I find it amusing when advisors, poobahs, and former middle school teachers tell another person to innovate. Yep, that works really well. Even those who innovate are faced with failure many times. I think the well ran dry for some of the Italian Renaissance artists when the examples of frescos in Nero’s modest home were recycled. Been there. Done that. The notion of a person innovating with an enterprise deployment of Hadoop strikes me as interesting, but probably not a top 10 priority. How about getting the data into the system, formulating a meaningful query, and figuring out how to deal with the batchiness of the system?

Item 9 of 10: “Look for capabilities that make Hadoop data look relational.” There is a reason to use Codd type data management systems. Those reasons include that they work when properly set up, and they require data which can be sliced and diced. Maybe not easily, but no one fools himself or herself thinking, “Gee, why don’t I dump everything into one big data lake and pull out the big, glossy fish automagically.”

I am okay with advice. Perhaps it should reflect the reality with which open source data management tools present to an enterprise user seeking guidance. Enterprise search vendors got themselves into a world of hurt with this type of casual advice. Where are those vendors now?

Stephen E Arnold, November 12, 2015

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta