This document describes how to deploy Hadoop clusters on OpenPOWER systems using OpenStack and the Sahara plugin in 3 steps: 1) Setup OpenStack with Sahara on OpenPOWER servers, 2) Create PowerPC images and node group templates in Sahara, 3) Launch and test a Hadoop cluster from the Sahara dashboard. The deployment was tested on IBM S822L servers running PowerKVM with a 500GB Terasort completing in 7000 seconds on 2 data nodes and 1 name node. Upstream contributions were also made to OpenStack to support PowerPC.

1. One-click Hadoop Cluster
Deployment on OpenPOWER
Systems
Pradeep K Surisetty
IBM
#OpenPOWERSummit

2. © 2014 IBM Corporation
2
#Whoami l
Systems & Infrastructure Engineer
l
9 Years + of Linux, Virtualization
l
Believe in Open Source Everything
l
Virtualization Test Lead/Solution Engineer
l
pradeepkumars@in.ibm.com
This is a team work:
Core Team:
• Pradipta Kumar, Pradeep K Surisetty, Ashish Kumar, Yogananth Subramaniyan, Poornima
Nayak, Sudeesh John
Acknowledgements:
• Dipankar Sarma, Vaidyanathan Srinivasan, Tarundeep S Kalra, Anbazhagan Mani, Ashish
Billore, Akash Gunjal
Join the conversation at #OpenPOWERSummit

3. © 2014 IBM Corporation
3
Elastic Hadoop on OpenPower Systems
Goal
●
Make Deployment & Operation of Hadoop Clusters simple on OpenPower Systems
●
Managed by OpenStack.
●
Run Hadoop Performance Benchmarks on this cluster. .
Key characteristics
●
Opensource Hadoop
●
OpenStack Native
●
Example OpenStack Sahara based Elastic Hadoop on OpenPower Servers
●
Benchmark Results
Join the conversation at #OpenPOWERSummit

4. © 2014 IBM Corporation
4
Intro on OpenStack and Sahara
OpenStack core components
Compute - Nova
Networking -Neutron
Object Storage – Swift
Block Storage – Cinder
Dashboard - Horizon
Identity Service - Keystone
Image Service - Glance
OpenStack Sahara Project
Sahara project is an initiative to provision
Hadoop on top of OpenStack (started by
Mirantis, Hortonworks and Red Hat)
Join the conversation at #OpenPOWERSummit

5. © 2014 IBM Corporation
5
High Level Architecture Overview
VM 1 VM 1VM 2 VM 2
PowerKVM
Compute Node
PowerKVM
Compute Node
HDFS
Openstack
Controller Node
Local Disk
Compute + Storage Node
●
Nova-compute
●
Cinder-volume
Controller Node
●
Nova-api
●
Cinder-scheduler
●
Cinder-api
●
Glance
●
Neutron
●
Horizon
●
Sahara
Join the conversation at #OpenPOWERSummit

6. © 2014 IBM Corporation
6
Test Environment Details
Hypervisor
Version PowerKVM-2.1.1
Kernel 3.10.42 -2015.1.pkvm2_1_1.40.
VM
OS RHEL7 PPC64
Kernel 3.10.0-123.el7
VCPU 8
Memory 40G
OpenStack
Version Juno
Sahara Upstream
Diskimage-builder Upstream
Infrastructure
Hardware IBM S822L
Socket 2
CPU 12
Memory 1TB
Disk 7.2TB
RAID 0
Hadoop Cluster
Hadoop 2.5.2
Data Node 2
Name node 1
Join the conversation at #OpenPOWERSummit

7. © 2014 IBM Corporation
7
Elastic Hadoop Work flow
1. Setup OpenStack Controller with Sahara plugin
2. Add Power/KVM compute nodes to OpenStack controller
3. Create Power arch (ppc64) images for Sahara
sahara-image-elements/diskimage-create/diskimage-create.sh -p vanilla -v 2.4 -i fedora
4. Register Image with Sahara
5. Create Node Group Templates based on required processes in the nodes.
●
Worker Template having only Data Node
●
Master Template Having Name node, Resource Manager, Node Manager
6 . Create Cluster Template as required
7. Launch Cluster based on template
8. Submit jobs to the Cluster
Demo Video: https://www.youtube.com/watch?v=JMprhJAF8FQ
Join the conversation at #OpenPOWERSummit

8. © 2014 IBM Corporation
8
Performance Results
Terasort for 500 GB of workload took 7000 seconds on this environment
with 2 Data nodes, 1 Name node
Join the conversation at #OpenPOWERSummit

9. © 2014 IBM Corporation
9
Upstream Contributions for PPC64 support
https://review.openstack.org/#/c/149045/
https://review.openstack.org/#/c/149165/
https://review.openstack.org/#/c/153404/
3. Enable vm element to create PowerPC image
2. Add support for using local PowerPC VM image
1. Ramdisk-image-create: Add support for vmlinux file
Join the conversation at #OpenPOWERSummit

10. © 2014 IBM Corporation
10
Upstream Contributions for PPC64 support
https://review.openstack.org/#/c/149045/
https://review.openstack.org/#/c/149165/
https://review.openstack.org/#/c/153404/
3. Enable vm element to create PowerPC image
2. Add support for using local PowerPC VM image
1. Ramdisk-image-create: Add support for vmlinux file
Join the conversation at #OpenPOWERSummit

11. © 2014 IBM Corporation
11
Reference & Demo Video

Hadoop on PowerKVM video:
https://www.youtube.com/watch?v=JMprhJAF8FQ

Creating an OpenStack cloud using DevStack and Power8 Compute Nodes
http://goo.gl/ZHYsot

Creating Openstack cloud using IBM Cloud Manager and Power8 Compute Nodes
http://goo.gl/3f46Lv

Hadoop Releases:
http://goo.gl/MOTq1x
http://hadoop.apache.org/releases.html/
Join the conversation at #OpenPOWERSummit

12. © 2014 IBM Corporation
12
Summary
• Hadoop Deployment & Operation can be done seamlessly on
OpenPower systems using OpenStack and Sahara
Join the conversation at #OpenPOWERSummit

13. © 2014 IBM Corporation
13Join the conversation at #OpenPOWERSummit
Post your questions here
pradeepkumars@in.ibm.com
bpradipta@in.ibm.com