How much Java is needed to use Hadoop and Mahout effectively?
I am a PHP developer. Let’s figure it out now. But Hadoop – and Mahout in particular – piqued my interest. I’m ready to dive into Java in order to use them.
So based on enough experience people can know, how much Java do I need to use them effectively? As far as I’ve seen, programming the mapper/reducer doesn’t solve all these problems a lot. But with Mahout, when I looked at the docs, I wasn’t at all sure what I was looking at.
Also, how hard would it be to get data from my PHP application via Hadoop and Mahout to process in Java? I can’t imagine it going to be that difficult, but I’m inexperienced enough to say.
Solution
Using Mahout and Hadoop to fetch data from PHP to Java for analysis shouldn’t be that difficult.
It’s simpler to use Mahout and Hadoop to process offline in batches and store data products in a file system or database. PHP can then read these data products as easily as they fall off the log.
For real-time use, Mahout’s recommended section supports various web service interfaces that can be accessed very easily from PHP. Using the model evaluation part of Mahout requires more programming.