I'm a PHP developer. Let's just get that out of the way now. But Hadoop – and Mahout in particular – have piqued my interest. I'm ready to take the dive into Java in order to use them.
So from people experience enough to know, just how much Java will I need to be able to use these effectively? From what I've seen, programming mappers/reducers doesn't take all that much. But with Mahout I'm not at all sure what I'm looking at when I look at the documentation.
Also, just how hard will it be to take data from my PHP application for processing in Java via Hadoop and Mahout? I can't imagine it'd be that difficult, but I'm not experienced enough to say.
It shouldn't be all that difficult to get data from PHP to Java for analysis using Mahout and Hadoop.
Even easier is to process using Mahout and Hadoop off-line in a batch mode and to store the data products in a file system or database. PHP can then read these data products as easy as falling off a log.
For real-time use, the recommendations part of Mahout supports a variety of web-service interfaces that make it pretty easy to access from PHP. Hitting the model evaluation part of Mahout would require a bit more programming.