With the collaboration of Hortonworks, Microsoft has made Hadoop framework work with Windows Servers and Windows Azure, and they called it HDInsight.
For those who are unfamiliar to Hadoop, you can think of Hadoop as a Framework if you want to make a very short definition. Under this Framework we have MapReduce (data processing technology) and HDFS (Hadoop File System) and various projects (such as Pig) as the basis.
The client is requesting information from Hadoop, the Master in the Master-Slave structure meets the demand in the Cluster and divides the data processing between the slaves and the processed data goes directly from the Slaves to the Client.
The main goal with Hadoop is to process a large amount of data (big data) in parallel (by slaves) in a scalable system (performance increases linearly as slave is added to the cluster).
Note: My friends who are curious about Hadoop can get a lot of videos from Youtube on the subject for more details.