Installation Guide

Prerequisites

All dependencies and their installation is managed via Ansible which we run through a docker container. So, the only required dependency for this setup is docker. Follow the steps specified at this link and then allow non-root users to use docker by following these steps.

Testbed Setup

The physical resources required on each machine are shown in the following figure.

The testbed requires 3 machines:

• COLLECTOR → to store rulesets/traces and collect logs
• GVS → device-under-test to run gvs
• TGEN → to send/receive traffic

Hardware Requirements

• CPU: Intel Xeon Platinum 8358P (64 cores) @ 2.60GHz
• Memory: 512 GB RAM
• SmartNIC
  • Intel XL710 (10/40G), Dual-port (for software-only Gigaflow)
  • Xilinx Alveo U250 FPGA (hardware offload, support coming soon!)

Workloads

To evaluate a virtual switch, you need some packet pipeline rulesets and matching traffic traces. We have provided a set of 5 real-world vSwitch pipelines, their corresponding rulesets and traffic traces that we used for benchmarking Gigaflow. The following are their detailed descriptions (more details in the paper).

Download

Example pipelines, their rulesets, and traffic traces used for benchmarking Gigaflow are publicly available via FigShare. Download and place them on the COLLECTOR machine as following:

Step 1: create directory

mkdir ~/Gigaflow
cd ~/Gigaflow

Step 2: download the traffic traces and pipelines

# download the CAIDA traffic traces
wget --content-disposition "https://figshare.com/ndownloader/files/52608875"
# .. and vSwitch pipelines
wget --content-disposition "https://figshare.com/ndownloader/files/52608872"

Step 3: unzip the downloaded files

unzip Traffic-Locality.zip
unzip vSwitch-Pipelines.zip

Step 4: rename downloaded directories to match the vars/main.yml file

mv Traffic-Locality mini_profiles
mv vSwitch-Pipelines ovs-pipelines

Now, we are ready to setup the orchestrator and install gvs and the traffic generator (tgen).

Orchestrator Setup

You only need to setup the gigaflow-orchestrator repository that will bringup the testbed, install all dependencies (including gvs and traffic generator), and run the experiments. The orchestration is enabled via Ansible which itself is provided as a docker container.

Clone the gigaflow-orchestrator repository as following:

git clone https://github.com/gigaflow-vswitch/gigaflow-orchestrator.git

Update Local Paths

In this repository, modify the following variables in the vars/main.yml file:

171
172
173
174
175
176
177

retrieve:
  caida:
    path: "/home/<username>/Gigaflow/mini_profiles"
  pipelines:
    path: "/home/<username>/Gigaflow/ovs-pipelines"
  destination: 
    path: "/tmp/{{ project.name }}/pipelines-and-traffic"

Update only the retrieve.caida.path and retrieve.pipelines.path variables to point to a Gigaflow directory on the COLLECTOR machine.

Inventory Configurations

We use Ansible to orcherstrate all experiments using the three machines. Therefore, we require root access to each of them. To populate for each machine, update the inventory.ini file as following:

1
2
3
4
5
6

[NODES]
TGEN ansible_host=<tgen-ip> ansible_user=<tgen-username> ansible_password=<tgen-password> ansible_sudo_pass=<tgen-root-password>
GVS ansible_host=<ovs-ip> ansible_user=<ovs-username> ansible_password=<ovs-password> ansible_sudo_pass=<ovs-root-password>

[STORAGE]
COLLECTOR ansible_host=<collector-ip> ansible_user=<collector-username> ansible_password=<collector-password> ansible_sudo_pass=<collector-root-password> ansible_ssh_user=<collector-username> ansible_ssh_pass=<collector-root-password>

To test if all machines are reachable, run the following command:

cd gigaflow-orchestrator
make ansible

This should start an Ansible docker container. Run the next commands from inside this container.

make ping

This should be successful and return something like this:

root@nga2-vm2:/workdir# make ping
ansible all -m ping
COLLECTOR | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python3"
    },
    "changed": false,
    "ping": "pong"
}
GVS | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python3"
    },
    "changed": false,
    "ping": "pong"
}
TGEN | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python3"
    },
    "changed": false,
    "ping": "pong"
}

This means all machines are reachable and you can now proceed to the next step.

GvS and TGen Installation

Now, you can install gvs and tgen along with all their dependencies as following:

make setup-gvs-experiment

You can also install them separately as following:

make install-dataset
make install-gvs
make install-tgen

At this point, you have succesfully installed gvs on the GVS machine and tgen on the TGEN machine. You have also retrieved and placed your pipelines and traffic traces on both machines and now you are ready to start running experiments!

Directory Structure

gigaflow-orchestrator/
├── roles/          # Ansible roles for all components needed to run GvS experiments
│   ├── collector/
│   ├── dpdk/
│   ├── gvs/
│   ├── logging/
│   ├── retrieve/
│   ├── rules/
│   └── tgen/
├── scripts/        
├── vars/           # Experiment variables
├── inventory.ini   # Ansible inventory file
├── ansible.cfg     # Ansible configuration file
├── Makefile        # Makefile for the ansible playbook targets
├── gvs.yml         # top-level gvs ansible playbook
├── tgen.yml        # top-level tgen ansible playbook
└── ...             # other top-level ansible playbooks

Next Steps

For detailed instructions on using Gigaflow in your testbed, please refer to our usage guide.

To evaluate Gigaflow performance against Megaflow cache with real-world workloads, please refer to our benchmarking guide.

Additional Documentation

• Getting Started
• Technical Details
• Performance Benchmarks
• Usage Instructions