Pls see this link, I am sure it will help you:

http://stackoverflow.com/questions/9365812/how-to-find-the-cpu-time-taken-by-a-map-reduce-task-in-hadoop

Thanks Reza.  One more question is, when I try the MapReduce program(wordcount or for any use case) from eclipse then though I dont start the ./start-all.sh, my program in eclipse works and gives me some output. Obviously, its not same as how I expected.

Now my question is, does it automatically start the hadoop using ./start-all.sh or without that it shows output when I exec it from eclipse?

Thanks in advance.

Running the MapReduce program through eclipse is used mostly for debugging. I am assuming that you are running it on your local machine, in which case you won't get the benefits of using a cloud. 

To answer your question, it really depends on the variables you setup in your Eclipse project. From what it sounds like, you either already have the service running (from a past execution on ./start-all) or you setup Eclipse such that it will start and run off local host automatically. ./start-all.sh simply boots up the mapreduce services (name-node, etc) so you can execute your programs. 

Hope this helped. Cheers.

I am worried about the CPU time mainly.  The output of the logs says the CPU time in seconds , unlimited. So do I have to convert or is there any setting  so that it will be calculating always in milliseconds. As every time , my CPU time has been shown as 0.

Thanks Ben! Yes , I did it like how you suggested. In addition to it, calculated total time taken by the mapper and reducer respectively and displayes it. Thanks to my colleagues who helped me!!