Planning

Platform requirements

Supported platforms for deployment

The following table lists the supported platforms when you are deploying a version of Maximo Visual Inspection Edge:

Table 1. Supported platforms

Version number	x86	IBM POWER8	IBM POWER9
8.9+	Supported	Not supported	Not supported
8.8	Supported	Not supported	Supported
8.7	Supported	Supported	Supported

Supported models for local deployment

A NVIDIA GPU is required for local deployment.

The following table lists the supported models when you are locally deploying a version of Maximo Visual Inspection Edge:

Table 2. Supported models for local deployment

Version number	Supported model
All versions	Faster R-CNN
All versions	SSD
All versions	GoogLeNet
8.5 +	YOLOv3
8.5 +	Tiny YOLOv3
8.6 +	Detectron2
8.7 +	High resolution
8.7 +	Anomaly optimized

Note: Action detection models are not supported for local deployment. Other types of models that are deployed in Maximo Visual Inspection can be run remotely in Maximo Visual Inspection Edge.

Hardware requirements

Supported GPU devices

For a list of supported GPU devices, see Supported GPU devices.

Disk space

• Installation

  The installation package contains Docker containers for deployment on all supported platforms. Images are installed in the file system that is used by Docker. The file system is usually /var/lib/docker.Only the images that are needed for the platform are installed.

• Deploying a model: 1 GB

  Models that are downloaded to Maximo Visual Inspection Edge are saved to a local directory that is volume-mapped to the Maximo Visual Inspection Edge Docker containers. Volume mapping ensures that when models are deployed on the Maximo Visual Inspection Edge server and are extracted inside the containers, the model artifacts are generated on the local file system.

• Metadata and image storage

  Storage requirements vary depending on the number of inspections, image inspection rates, image sizes, the number of detected objects in images, and the number of rules per inspection. For example, a system that continuously runs 10 inspections, which each process one image per second that has a resolution of 640 x 480, requires approximately 750 GB of storage per week.

Software requirements

Linux

• Ubuntu 20.04
• Red Hat® Enterprise Linux® 7.6 ALT for IBM POWER8
  Note: Starting in Maximo Visual Inspection Edge 8.9, IBM Power® Systems are no longer supported.
• Red Hat Enterprise Linux 7.7 for x86 and IBM POWER9
  Note: Starting in Maximo Visual Inspection Edge 8.8, IBM POWER8 is no longer supported. IBM POWER9 is the only supported Power Systems platform.

• Red Hat Enterprise Linux 8.

  Note: Support for Red Hat Enterprise Linux 8 starts in Maximo Visual Inspection Edge version 9.0.

NVIDIA GPU driver

• x86 servers with 11.8 or later drivers.
• Power Systems servers with 11.0 or later drivers.

For more information, see the NVIDIA CUDA Toolkit website.

NVIDIA Container Toolkit

For more information, see Installing NVIDIA Container Toolkit and Docker.

Containerization support

For containerization support, install either Podman or Docker.

Podman

Starting in Maximo Visual Inspection Edge version 9.0, the application runs on Red Hat Enterprise Linux 8.9 and requires the installation of Podman. For more information, see the Podman Installation Instructions on the Podman website.

For more information about installing Podman, see Installing Podman.

Docker

For both Red Hat Enterprise Linux 7 and Ubuntu, Maximo Visual Inspection Edge requires Docker Engine version 18 or greater to be installed.

When you run Ubuntu 20.04, nvidia-docker2 is supported.

For more information about Red Hat Enterprise Linux 7, see Using nvidia-docker2 with Red Hat Enterprise Linux 7.

Note: On Red Hat Enterprise Linux 8, Podman runs containers instead of Docker.

For more information about installing Docker and nvidia-docker2, see Installing Docker.

Storage

Storage optimization

When a Maximo Visual Inspection Edge system runs at full capacity, it produces a large volume of image files and metadata. Because the available compute resources on the system determine how much data is produced, no single storage configuration can be recommended for all deployments. However, you can take the following actions to optimize your storage configuration:

• Create the root installation directory on a different storage volume than the volume that Docker uses, which is normally /var.
• To prevent a single storage device failure from affecting both transactional data and database backups, ensure that the following database directories are located on different volumes:
  • <installation root>/volume/run/psdata
  • <installation root>/volume/run/pstbspc
  • <installation root>/volume/run/pgbackrest

GPU memory

Models that you download and deploy in Maximo Visual Inspection Edge are deployed across all GPUs in the system. The Models includes a GPU usage meter that displays the free and available memory for all GPUs. Before you deploy a model, check the meter and ensure that your system can accommodate the model.

MQTT

Maximo Visual Inspection Edge requires an MQTT broker, such as IBM Integration Bus, App Connect Enterprise Enterprise, HiveMQ, or Mosquitto, to be configured to enable external systems to communicate with the application through MQTT topics. Inbound messages from external systems trigger cameras to take photos and outbound messages from Maximo Visual Inspection Edge send alert messages to recipients to notify them of inspection results and required actions.

Maximo Visual Inspection Edge includes an internal instance of the Mosquitto MQTT broker that enables external systems to communicate with the application through MQTT topics. When an MQTT client publishes a message in a topic, the broker sends the message to all of the clients who subscribe to that topic.

Inbound messages from external systems that trigger input sources, such as cameras, cameras, to capture images are sent to the internal MQTT broker. If systems in your workflow already publish trigger messages to an external broker and you want Maximo Visual Inspection Edge to subscribe to those topics, MQTT bridging must be configured between the external and internal brokers.

Outbound messages from Maximo Visual Inspection Edge send alerts to external recipients, such as supervisors, to notify them of inspection results and required actions. Recipients can subscribe to the internal broker to receive outbound messages. Optionally, you can configure MQTT settings in Maximo Visual Inspection Edge to send outbound messages to an external broker. See MQTT configuration for further information.

Twilio

A Twilio account must be set up to enable alert messages to be sent to external recipients. See the Twilio documentation for further information.

Supported web browsers

The following web browsers are supported:

• Google Chrome Version 60 or later
• Firefox Quantum 59.0 or later

Limitations

The maximum number of input sources, such as IP cameras, that you can add in Maximo Visual Inspection Edge is 10 for each GPU in your hardware system.

Installation