Skip to content

Deployment Scenarios

The Starter Kit support a wide variety of deployments, and this article will highlight some of them.

Other scenarios are supported, and combinations of the below scenarios are possible.

Keep data on-premise

It is possible to deploy Azure IoT Edge on your own hardware, and keep all data in your local network or infrastructure. For example, by using a custom local forwarding module, it is possible to route the data to a local queue or message bus. In this case, the sensor data will never leave the on-premise network, and the connection to Azure will be used for managing the devices, handling the deduplication or managing concentrator updates for example.

Keep data on-premise

Scalability & Availability

There are multiple strategies that you can take to improve the availability of the starter kit listed below. For the scenarios that use multiple LNS, you can further increase the availability of a concentrator by using a highly available LNS discovery service.

Note: This guide is only talking about the LBS/LNS deployment configuration. For the full solution, the function endpoint, IoT Hub and Redis need to be taken into consideration as well. There is documentation for all of these services explaining the different options for HA and scalability.

With the version 2.1.0 we introduced a major improvement for multi gateway deployments that allows the LNS to scale much better when using DeduplicationMode.Drop. All the deployment scenarios are now scaling to similiar levels no matter if a multi gateway deployment is used or not.

Single concentrator with single LNS

This is the most basic deployment you can chose. It does only deploy 1 concentrator and one network server. Each sensor can at most reach a single concentrator.

Advantages

  1. Simple: the deployment is simple to start with and maintain.
  2. Low cost: No redundant hardware is used, which helps keeping the operational cost down.

Disadvantages

  1. Availability: this solution has 2 points of failure: the concentrator and the network server. Either of those can go down. The result in the worst case is loss of messages.
  2. Limited reach: Since there is only a single concentrator, you have to put all your sensor within reach of that concentrator.
  3. Reliability: message delivery might be impacted.
Target Setting Value
Device Twin KeepAliveTimeout undefined or 0
Device Twin Deduplication None
Device Twin GatewayId The Id of the Network Server
Edge Hub environment variable MaxConnectedClients Depends on number of devices
Network server environment variable IOTHUB_CONNECTION_POOL_SIZE Depends on the number of devices

Single concentrator - Single gateway

Multi concentrator with single LNS

Multiple concentrators are deployed that reach a single network server. Each sensor can reach one or multiple concentrators. Depending on the goal of the deployment (availability and/or reach).

Advantages

  1. Relatively simple
  2. Low cost
  3. Partial redundancy: if the deployment is designed to ensure that each sensor can reach at least 2 concentrators, you get higher availability than with single/single.
  4. Reach: if you deploy for reach, you can get a larger support a larger geographical deployment. To combine reach with availability, the cost will raise.

Disadvantages

  1. Availability: Depending on the deployment, this solution has at least one single point of failure: the network server. The result in the worst case is loss of messages.
  2. Reliability: message delivery might be impacted.
Target Setting Value
Device Twin KeepAliveTimeout undefined or 0
Device Twin Deduplication None
Device Twin GatewayId The Id of the Network Server
Edge Hub Environment variable MaxConnectedClients Depends on number of devices
Network server environment variable IOTHUB_CONNECTION_POOL_SIZE Depends on the number of devices

Multi Single

Multiple concentrator with Multiple LNS

Multiple concentrators are deployed that reach multiple LNS (note one concentrator can only be connected to a single LNS). Each sensor can reach one or multiple concentrators. Depending on the goal of the deployment (availability and/or reach).

To get high availability concentrator deployment and assignment to LNS should be done so that a single sensor is in range of at least 2 concentrators, both connecting to a different LNS.

Advantages

  1. Availability: If the deployment is done right, the single point of failure can be eliminated.
  2. Reliability: message delivery likelihood is increased.
  3. Reach: The deployment model can be expanded

Disadvantages

  1. Cost: all components need to be deployed and managed multiple times.
  2. Complexity
Target Setting Value
Device Twin KeepAliveTimeout > 0 - depending on the sensor, this can be tuned to be the time you expect a new message to come in. Ideally not > 300
Device Twin Deduplication Drop
Device Twin GatewayId empty

Redundancy

Deployment on Windows Server

Some organizations only support Windows Server operating systems on their infrastructure. In this case, it is still possible to use IoT Edge and the Starter Kit by using EFLOW, short for "Azure IoT Edge for Linux on Windows". This allows you to run containerized Linux workloads on Windows machines.

Redundancy

Deployment of IoT Edge in the cloud

It is also possible to run the IoT Edge modules directly in the cloud. Either a Linux VM to run IoT Edge directly, or on a Windows VM with EFLOW. In fact, the full end-to-end Continuous Integration pipeline for the Starter Kit also tests this scenario with an installation of EFLOW in a Windows Server VM in Azure.

Redundancy

Appendix

Last update: 2022-03-02
Created: 2022-01-26
Back to top