Back to Top

Physical architecture

This architecture was designed with:

  • A small/medium-sized gas utility in mind
  • Support for workflows with a target design load of 15 ArcGIS Pro editors and 200 ArcGIS web users (general user personas).
  • An enterprise geodatabase configured with Microsoft SQL Server
  • AWS cloud infrastructure

The system was designed for and tested against the specified workflows, with adjustments to the machine types and sizes as needed based on test results.

gas-sql-refarch.png

Note:

For more information on the software components and key interactions in this architecture, please review the Network Management System reference architecture.

Machine types and sizes

The following represent the machine sizes chosen and validated for the scope and purpose of this test study. However, it is highly recommended to follow a complete design process to account for your business and technical requirements.

Esri offers system architecture design services should you need help determining all of the different factors relating to your organization’s physical design- such as networking, storage, system environments, and sizing. Minimum system requirements for each component are listed in the respective software documentation available online.

Desktop (ArcGIS Pro & web browser)

  • 3 machines (used in testing
  • g4DN.2xlarge instance type
  • 4 CPU (8 vCPU)
  • 32 GB RAM
  • 16 GB GPU
  • 1 TB Disk

Portal for ArcGIS

  • 2 machines
  • m6i.xlarge instance type
  • 2 CPU (4 vCPU)
  • 16 GB RAM
  • 128 GB Disk

ArcGIS GIS Server

  • 2 machines
  • m6i.2xlarge instance type
  • 4 CPU (8 vCPU)
  • 32 GB RAM
  • 128 GB Disk

ArcGIS Server (hosting server)

  • 2 machines
  • m6i.2xlarge instance type
  • 4 CPU (8 vCPU)
  • 32 GB RAM
  • 128 GB Disk

ArcGIS Data Store (relational)

  • 2 machines
  • m6i.xlarge instance type
  • 2 CPU (4 vCPU)
  • 16 GB RAM
  • 256 GB Disk

ArcGIS Web Adaptor

  • 2 machines
  • m6i.large instance type
  • 1 CPU (2 vCPU)
  • 8GB RAM
  • 128 GB Disk

ArcGIS Monitor

  • 1 machine
  • m6i.2xlarge instance type
  • 4 CPU (8 vCPU)
  • 32 GB RAM
  • 128 GB Disk

File storage

  • 1 instance
  • 64 MB/s throughput
  • 1 TB Disk

Database

  • 1 machine
  • m6i.4xlarge instance type
  • 8 CPU (16 vCPU)
  • 32 GB RAM
  • 128 GB Disk

Domain server

  • 1 machine
  • c6i.large instance type
  • 1 CPU (2 vCPU)
  • 4 GB RAM
  • 128 GB Disk

Additional infrastructure considerations

The following are additional areas of consideration when designing a Network Management System and an explanation of some infrastructure choices made for this test study.

Application load balancer (ALB)

At least one third-party load balancer is required in a highly available ArcGIS Enterprise deployment- they handle client traffic to your portal and server sites and internal traffic between the software components. Although the ArcGIS Web Adaptor is considered a load balancer, it is insufficient on its own to serve as a load balancer in a high availability configuration. Therefore, in this test study, an Amazon Web Services (AWS) Application Load Balancer was used.

Database consideration

For the scope and purposes of this test study, a choice was made to deploy SQL Server on a virtual machine (EC2). However, depending on your needs, you might consider leveraging a database platform service like RDS (Amazon Relational Database Service) or SQL Managed Instance in a Microsoft Azure environment.

Shared storage

To successfully implement a highly available ArcGIS Enterprise deployment, the configuration stores must be stored in a highly available, shared location. This ensures this data remains accessible even if one server fails, providing uninterrupted service to end-users. Additionally, shared storage simplifies data management in a multi-machine deployment and improves scalability by centralizing data storage and allowing for expansion as needed.

System components not included in the diagram

Note that while antivirus software and AWS networking components are not illustrated in this diagram, they were present in the test study.

Top