Hardware Sizing

Introduction

This document provides comprehensive guidance on estimating the hardware requirements for running N3uron. Designed to be lightweight, N3uron is capable of managing thousands to tens of thousands of tags on resource-constrained devices like a Raspberry Pi. With more powerful hardware, N3uron scales seamlessly, enabling it to handle hundreds of thousands of tags on a single node, delivering exceptional performance for a wide range of deployment scenarios.

To ensure optimal performance and scalability, the following factors should be considered when estimating hardware requirements:

• Number of Tags: Memory usage increases linearly with the number of tags.

• Event Rate (tag changes per second): Higher event rates demand more CPU resources. Adjusting scan rates and deadband settings helps optimize performance.

• Modules in Use: Each module instance creates a new process, impacting systems with limited memory.

• Scripting and Edge Computing: Complex scripts and long-running calculations can significantly increase CPU and memory usage.

• Data Historization: Running the Historian database on the same node increases memory usage, requires fast I/O, and reduces event-handling capacity as system resources are shared between N3uron and the database.

• Web Vision Design: The design of Web Vision screens, including the number of visualized tags, historical data queries, and refresh rates, impacts system performance.

Note:

At the hardware level, high-frequency CPU cores and fast SSD storage ensure the best performance for N3uron.

Historian Database

The N3uron Historian is a high-performance, space-efficient time-series data historian built on MongoDB.It is optimized to handle large amounts of data while maintaining minimal storage requirements. For the most common N3uron workloads, the Historian requires ≈10 bytes/event, meaning it can store 100,000,000 events using 1 GB of disk space.

To estimate the disk space required for your workload, consider the number of events per second (Event Rate) to be historized and the duration for which the data will be stored (Retention Period).

Here are some examples:

Note:

Regarding disk type and speed, we strongly recommend using SSD storage for the Historian database to ensure optimal performance and reliability. For workloads with very high event rates, it is crucial to provision fast SSDs with high throughput and IOPS (Input/Output Operations Per Second).

Edge Deployments

Note

Please make sure you test and validate your solution based on your workload, using the recommended hardware as a reference.

Small Deployments

Recommended hardware for N3uron deployments in small plants, handling a low number of signals and events per second, primarily focused on data collection with limited historization, basic computations, and data exchange with other systems.

4 CPU cores (ARM), 1GB RAM

This is a low-power ARM device, similar to a Raspberry Pi 3/4 or equivalent.

Without a local Historian:

10,000 tags

~2,000 events per second

With a local Historian:

5,000 tags

~1,000 events per second

4 CPU cores (ARM), 2GB RAM

This is a low-power ARM device, similar to a Raspberry Pi 3/4 or equivalent.

Without a local Historian:

40,000 tags

~2,000 events per second

With a local Historian:

20,000 tags

~1,000 events per second

1 CPU core (4GHz), 1GB RAM

Without a local Historian:

10,000 tags

~5,000 events per second

With a local Historian:

5,000 tags

~1,500 events per second

1 CPU core (4GHz), 2GB RAM

Without a local Historian:

40,000 tags

~5,000 events per second

With a local Historian:

20,000 tags

~1,500 events per second

Medium

Recommended hardware for N3uron deployments in medium-sized plants, capable of handling a larger number of signals, performing more complex edge computing, data historization, and providing basic data visualization.

4 CPU core (ARM), 8GB RAM

This is a more performant ARM device, similar to a Raspberry Pi 5 or equivalent.

Without a local Historian:

100,000 tags

~6,000 events per second

With a local Historian:

100,000 tags

~3,000 events per second

2 CPU cores (4GHz), 4GB RAM

Without a local Historian:

70,000 tags

~20,000 events per second

With a local Historian:

50,000 tags

~7,000 events per second

2 CPU cores (4GHz), 8GB RAM

Without a local Historian:

150,000 tags

~20,000 events per second

With a local Historian:

150,000 tags

~10,000 events per second

Large

Recommended hardware for N3uron deployments in large plants, designed to manage the most demanding workloads, including a high volume of signals and events per second, robust data historization, and advanced visualization capabilities.

4 CPU cores (4GHz), 8GB RAM

Without a local Historian:

200,000 tags

~50,000 events per second

With a local Historian:

200,000 tags

~20,000 events per second

4 CPU cores (4GHz), 16GB RAM

Without a local Historian:

500,000 tags

~60,000 events per second

With a local Historian:

500,000 tags

~30,000 events per second

8 CPU cores (4GHz), 16GB RAM

Without a local Historian:

500,000 tags

~100,000 events per second

With a local Historian:

500,000 tags

~50,000 events per second

Cloud Deployments

Note

Please make sure you test and validate your solution based on your workload, using the recommended hardware as a reference.

The recommended architecture for deploying N3uron on a central server, whether in the cloud or on-premise, is to use at least two machines: one dedicated to running N3uron and another exclusively for the Historian database.

The examples provided below use Amazon Web Services (AWS) AMIs as a reference for the N3uron node, but equivalent configurations can be selected on other cloud platforms.

Small

2 vCPU, 8GB RAM

This corresponds to the AWS m7g.large or equivalent instance.

200,000 tags

~30,000 events per second

Medium

4 vCPU, 16GB RAM

This corresponds to the AWS m7g.xlarge or equivalent instance.

500,000 tags

~50,000 events per second

Large

8 vCPU, 32GB RAM

This corresponds to the AWS m7g.2xlarge or equivalent instance.

1,000,000 tags

~100,000 events per second