Fully automated benchmark for Fedora 4 Repository testing realistic workflows leveraging collected sensor data obtained from sensor- and accelerometer-outfitted Goodwin Hall at Virginia Tech. This raw data is stored in HDF5 files.
VT Fedora Benchmark is primarily meant to run withing AWS ecosystem. When evaluating Fedora using the benchmark overall workflow can be split into 2 parts:
- Setup - involves provisioning and subsequent provisioning of AWS resources. This part is fully automated using Ansible playbook.
- Execution - involves running workflows against resources provisioned in the step 1. This part is fully automated using Gradle build system.
After cloning the repository make sure to install and properly configure Ansible on your host machine. Please follow instructions mentioned in the Ansible documentation.
- Install Ansible - Installation and Getting Started sections
- Configure Ansible with EC2 - AWS EC2 External Inventory section
Note: make sure SSH key is set up either in the SSH Agent or by using the "--private-key" option to the "ansible" command invocation.
After Ansible has been installed and properly set up it is necessary to make sure AWS is properly prepared. The benchmark requires input HDF5 files to be available for processing in AWS S3 bucket. Please make sure to create personal S3 bucket and upload all the files you wish to use for processing.
Sample input set containing 60 HDF5 files of size 50MB is available here: https://drive.google.com/drive/folders/0B5gbHbvEraLnLXZjc055TUNpcW8?usp=sharing
In addition it is necessary to manually create EC2 IAM role named FedoraAnsibleS3ReadRole with at least
AmazonS3ReadOnlyAccess managed policy (you can specify AmazonS3FullAccess but it is not necessary).
Please note that not all versions of Ansible EC2 allow programmatically create IAM roles on AWS, which is
why, for now, is this Ansible playbook step commented out. If you're positive that your version of Ansible supports mentioned functionality feel free to uncomment following section in automation/aws-playbook-yml and skip this step.
After above-mentioned configuration has been performed you are ready to run the playbook and get resources
set up. All the necessary files are stored under automation directory.
- Setup Ansible global variables stored in automation/group_vars/all. Make sure to follow guidelines stored in the same file for each variable.
- Execute Ansible playbook - automation/aws-vm-playbook.yml
# example execution
ansible-playbook -i ec2.py automation/aws-vm-playbook.yml
Assuming that everything went well you now have configured AWS environment with all the major components installed and communicating with each other. Ansible setup additionally takes care of altering your local benchmark configuration file with correct values from AWS to enable smoother execution. The benchmark itself is mere Gradle-based Java application and thus all the Gradle commands apply.
Please note that
you don't have to have Gradle installed - the solution comes with 2 scripts - gradlew for Linux/OS X and
gradlew.bat for Windows. Feel free to replace below gradle commands with appropriate script.
gradle assemble
or
gradle build
# running in batch/offline mode
gradle run -Dadministrator=edu.vt.sil.administrator.BatchAdministrator
# running in interactive mode (read below for explanation)
gradle run -Dadministrator=edu.vt.sil.administrator.InteractiveAdministrator -Dexec.args="<enter arguments here>"
Above-mentioned guide allows you to get going as quickly as possible. If you wish to learn more about overall workflow different execution modes and experiment customization please visit following website: Fedora 4 Benchmarking