Hadoop Dont’s: What not to do to harvest Hadoop’s full potential by Iwona Bialynicka-Birula.
From the post:
We’ve all heard this story. All was fine until one day your boss heard somewhere that Hadoop and No-SQL are the new black and mandated that the whole company switch over whatever it was doing to the Hadoop et al. technology stack, because that’s the only way to get your solution to scale to web proportions while maintaining reliability and efficiency.
So you threw away your old relational database back end and maybe all or part of your middle tier code, bought a couple of books, and after a few days of swearing got your first MapReduce jobs running. But as you finished re-implementing your entire solution, you found that not only is the system way less efficient than the old one, but it’s not even scalable or reliable and your meetings are starting more and more to resemble the Hadoop Downfall parody.
An excellent post on problems to avoid with Hadoop!