ArcGIS Data Store: Dedicated Server vs Shared with ArcGIS Server for 3D Data

MaximeDemers

Hi everyone,

I'm currently planning an ArcGIS Enterprise deployment and would like your input on a key architectural decision.

Is there a significant advantage to installing ArcGIS Data Store on a dedicated server rather than on the same machine as ArcGIS Server, particularly when the Data Store will be used primarily for 3D data visualization (e.g., mesh layers, multipatch features, point clouds, etc.)?

We’re estimating around 2 TB of storage for the hosted data. Performance and scalability are important considerations, especially for smooth rendering and interaction with 3D content in web scenes and apps.

We would prefer the simplicity of installing ArcGIS Data Store on the same server as ArcGIS Server. However, if it's recommended or if there are clear advantages to using a dedicated server, we’re open to going that route.

If anyone has experience with similar setups or insights into best practices for this kind of workload, I’d really appreciate your thoughts.

Thanks in advance!

TimoT

When planning an ArcGIS Enterprise deployment that includes 3D data, it's critical to consider the version of Enterprise you’re working with. This has a direct impact on your storage architectural options.

Starting with ArcGIS Enterprise 11.4, Esri has deprecated the tilecache datastore, and instead uses an object store to house 3D data for hosted scene services. In 11.3 and earlier, tilecache datastore was still valid and supported.

If you're using ArcGIS Enterprise 11.4+, you have two main options for deploying the object store:

1. Object Store on Cloud Storage (e.g. Azure Blob Storage, Amazon S3)
2. Object Store on a machine with ArcGIS Data Store installed

If you're deploying ArcGIS Enterprise in Azure or AWS, I recommend looking into using their respective cloud-native storage (Blob/S3) for your object store. This approach offers several key benefits:

• Higher scalability and elasticity
• No OS or patching maintenance
• Built-in redundancy and durability
• More cost-efficient — you pay for what you use

In contrast, deploying the object store on a VM:

• Generally more costly
• Requires more planning around machine size/specs, disk IOPS, storage architecture, etc.
• Involves additional maintenance overhead (OS and ArcGIS Enterprise patching + maintenance, disk management, backups, etc)
• While scalable, not as elastic; increasing disk storage or CPU/Memory resources requires shutdown of machine.

All that being said, it's generally not that bad to manage additional machine resources, especially if you're already managing other machines.

Now, if you decide to deploy the object store on a server/VM, then to circle back to your original question:

No, it's not recommended to install the object store (or any ArcGIS Data Store) on the same machine as your ArcGIS Server for Production environments with medium-heavy load, even moreso when dealing with 3D data where both publishing and visualization are resource intensive.

Co-locating these components increases the risk of:

• CPU and memory contention
• Disk I/O bottlenecks
• Reduced performance and/or stability

Even without 3D workloads, Esri explicitly advises separating ArcGIS Server and Data Store components to avoid resource contention:

Esri recommends that each type of data store... | Install ArcGIS Data Store—ArcGIS Enterprise

If you're interested in understanding more about why and when to consider separating ArcGIS Enterprise components or functions onto dedicated machine resources, it's worth exploring the concept of workload separation. This principle is foundational to improving both performance and reliability in ArcGIS Enterprise deployments.

Workload separation | ArcGIS Architecture Center