Configuring the instance

Navigate from the Google Cloud dashboard to Compute Engine and find VM Instances.

Click on Create instance button located at the top of the page.

Name and region

Specify the name of your instance and the designated region in which the instance should be hosted.

Machine configuration

GPU selection

In this section, switch the instance type from General purpose to GPUs. Then list from GPU type drop-down menu and select any of the listed GPUs.

CPU selection

You can list from available CPU + RAM combinations and select one or, preferably, switch to Custom to adjust the number of vCPUs and amount of RAM by yourself.

🚧

HW availability

• The availability of HW configurations is different across various regions. If your configuration is not available in the specified region, you will be notified and offered an alternative region.
• Always watch your quotas - the configuration page won't notify you whether the HW specs are within your quotas, e.g. it won't check whether you can create an instance with 64 vCPUs and four GPUs Tesla T4. HW specs and quotas are evaluated once you hit Create button.

📘

HW recommendation

We recommend starting your testing with a single NVIDIA Tesla T4, which is able to handle the load in most use cases and it has a wide availability across regions.

To perform basic tests with 1-2 streams, 8 vCPUs with 16 GB of RAM are enough in most use cases.

Still don't know what HW configuration is best suitable for you? Don't worry, our team will gladly help you out. Contact us at [email protected] and describe your workflow for us.

Boot disk

The recommended host OS for Live Transcoder is Ubuntu 20.04 LTS (x86/64). You can select it by clicking on Change button in the Boot disk section.

For hourly-based streaming (less than 40 hours a month), the capacity of 128 GB will suffice. For HLS streaming (you can check our storage calculator) or more extended streaming periods, you should increase the storage size to at least 256 GB.

🚧

Make sure to check the Delete boot disk in Advanced configuration options so your quotas for disk space won't get filled up by deleted instances.

Firewall

It is mandatory to allow HTTP traffic to be able to

• Access the GUI to the Live Transcoder
• Receive an HLS stream or host the HLS output

Allowing HTTPS is mandatory only if your HLS sources are on secured HTTP.

Advanced options

Networking

All the necessary networking/firewall rules are set for a particular network, usually labeled as default, before or after the instance is created. You will find an example of firewall rules later in the guide.

Management

Find the Automation section and paste the following script in it:

#!/bin/bash

# log file for nvidia-smi output executed in docker via NVIDIA container toolkit (NTC)
NTC_NVSMI_OUTPUT_FILE="nvidia-smi-docker.log"

# and log docker-compose version
DOCKER_COMPOSE_V="docker-compose-version.log"

# if these files exist, the system preparation has been performed
# do not continue if any of them exist (the preparation might have terminated abruptly)
if [ -f ${NTC_NVSMI_OUTPUT_FILE} ] || [ -f ${DOCKER_COMPOSE_V} ] ; then
	exit 0;
fi

# disable nouveau
sudo modprobe -r nouveau
echo "blacklist nouveau" | sudo tee /etc/modprobe.d/blacklist.conf

# update the system and install nvidia packages
sudo apt-get update -y
sudo apt install -y nvidia-headless-535 nvidia-utils-535 libnvidia-encode-535
sudo modprobe nvidia

# download and enable docker
curl https://get.docker.com | sh && sudo systemctl --now enable docker

# subscribe to NTC repo
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \
      && curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
      && curl -s -L https://nvidia.github.io/libnvidia-container/$distribution/libnvidia-container.list | \
            sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
            sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

# after repo is updated, continue by downloading NTC
sudo apt-get update -y
sudo apt-get install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker
sudo docker run --rm --runtime=nvidia --gpus all nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi > ${NTC_NVSMI_OUTPUT_FILE}

# setup docker-compose
sudo curl -L https://github.com/docker/compose/releases/latest/download/docker-compose-$(uname -s)-$(uname -m) -o /usr/local/bin/docker-compose
sudo chmod +x /usr/local/bin/docker-compose
docker-compose version > ${DOCKER_COMPOSE_V}

This script installs all necessary packages, docker, docker-compose, and NVIDIA container toolkit for your convenience.

🚧

Check the OS compatibility

Make sure you selected Ubuntu 20.04 LTS as an Operating system, as the script is being tested and maintained against this operating system only.

With this configuration, you can finally Create the VM Instance and continue by setting up Firewall rules.

Firewall rules

Navigate from the Google Cloud dashboard to Compute Engine and find VM Instances.

Select the Set up firewall rules option.

There is already a set of rules for your default network which allows you to freely send/receive data between the instances on this particular network.

To allow ingress and egress traffic with world, you have to click on Create firewall rule.

Next up:

• Type in the rule's Name
• (Optional) Associate the rule with a particular Network
• Select Direction of traffic
• Target tags to apply the rule only for certain VM instances in the network, or select All instances in the network from the drop-down menu
• Specify the Source/Destination IPv4 ranges, port range, and protocol

Note that you must specify these rules for ingress and egress traffic separately by creating 2 firewall rules.

📘

NDI, SRT and RTMP firewall rules

To allow a network traffic ingress or egress for these protocols, you have to enable traffic via specific TCP and UDP ports, e.g.:

• NDI - Allow TCP and UDP ports in range 5353-7999, as described in NDI documentation
• SRT (incl. RTP/UDP) - Allow specific range for UDP ports for your MPEG-TS streams in/out, e.g. port range 5000-6000
• RTMP - Allow TCP port 1935