SEHADOOP IMPLEMENTATION & EXPERIMENTS
We implemented the SEHadoop model using Hadoop version
2.0.1-alpha. SEHadoop ran on virtual machines (VMs)
for the experiments. The guest operating system was Ubuntu
11.04. All VMs ran on VMware ESXi 5.0.0. And VMware
ESXi was installed on IBM eServer Blade Center HS21s with
two Intel E5450 Xeon CPUs. Each CPU has 4 cores and the
clock rate of 3 GHz. The network card was Intel Corporation
82545EM Gigabit Ethernet Controller, and network line-speed
was 1 Gbit per second. Both SEHadoop and Hadoop were
tested in the same cloud set-up. We used seven VMs for the
experiments with each VM having 7 GB available disk space
and one CPU core. One VM was the master VM with 6 GB
memory. Five VMs were slave VMs with 7 GB memory each.
A separated 1 GB memory VM ran Kerberos. The last one VM
named Kerberos VM has 1 GB memory. The Name Node, the
Resource Manager, and the Job Client ran on the master VM.
One Data Node and one Node Manager ran on each slave VM.