MapReduce in C++
Since I can’t request the admistrator to configure any particular setting for the cluster nodes. I found the simplest way is to compile your code locally, use hadoop streaming, ship all you liked os-files with your mapper (compiled executable) together to the node. You could write a bash codes as the mapper, which config the lib-path, and wrap the exe inside it. Below are some websites talking about MapReduce in C++, which actually very nice, but may need more to configure.
JavaCV : nice for using opencv http://code.google.com/p/javacv/
Very nice example(Mapper with Second key): http://code.google.com/p/hadoop-stream-mapreduce/