Hadoop Speed

When using hadoop on the ics server, it takes about 10 minutes to run a test. We packaged it using the FatJar extension to eclipse. Are there any recommendations to make this run any faster? The example takes like 1 minute. Even when we use the example code in our project that we make it still takes 10 minutes. It takes like 3 minutes to say "map 0%..." and then another 3 minutes to do anything and show up as a job. Once the map starts it's extremely fast. But when it finishes it takes another 2 minutes to close the job.

Thanks for your help!