Running Hadoop on Mesos allows you to share cluster resources with other frameworks. By running Hadoop on Mesos, we can leverage the huge community and frameworks built on top of Hadoop, such as Giraph, Hama, HBase, Hive, and so on. Also, if we are already using Hadoop, by running it on Mesos, we can continue using tools built around the Hadoop ecosystem, but with improved resource utilization. Existing MapReduce code and tools will continue working with Hadoop on Mesos. With Hadoop on Mesos, we can also run multiple versions of Hadoop on the same Mesos cluster resources. Hadoop on the Mesos project implements a Mesos framework scheduler around Hadoop's JobTracker and wraps Hadoop executors to Mesos executors. The following figure shows you how Hadoop and Mesos integrate.
In the following figure, the master node is running JobTracker on the Mesos master node and some Mesos slaves are running TaskTrackers. Each TaskTracker is running some Map and Reduce task slots. Note that...