I'm currently installing ArcGIS Enterprise on AWS EC2 instance in order to use in our development enviroment, so wich EC2 instance do you recomend for this purpose? What are the minimun requirements in CPU numbers and RAM?
Solved! Go to Solution.
You can start with a m5.2xlarge for a silo deployment of ArcGIS Enterprise with Portal + Server + DataStore on the same machine, then increase to a large m5 series machine if workload requires more cpu and memory.
I would look over the ArcGIS Enterprise system requirements for Windows or Linux here - https://enterprise.arcgis.com/en/system-requirements/latest/windows/arcgis-enterprise-overall-system.... If it's a single machine deployment, it's pretty simple to modify the instance type after deployment, if your original selection doesn't meet your needs.
Advice, create an Amazon EventBridge rule using Lambda to stop/strat your dev instances when not in use to save some dough. This video can help - https://www.youtube.com/watch?v=tniZDP4PDz0
You can start with a m5.2xlarge for a silo deployment of ArcGIS Enterprise with Portal + Server + DataStore on the same machine, then increase to a large m5 series machine if workload requires more cpu and memory.
Hi Marcelo,
Just found your post here and perhaps you can advise me. In ArcGIS Pro 3.2.1, I have large models in ModelBuilder with 4-5 Python scripts embedded as tools and I am looking for an efficient AWS EC2 machine to run my models (way) faster.
So far, I tried an EC2 r6a.4xlarge and I am rather disappointed. A model running in 1h01 on my laptop took 1h15 on this virtual machine. It does not make sense as this r6a.4xlarge has 16 vCPU with 128 GiB vs 4 CPUs with 32 Gb.
In theory, I guess I don't need GPUs but would it help to select a GPU enabled virtual machine?
What Marcelo stated is going to be the best starting option. We have our GIS in AWS and we run m5.2xL with that stack (Portal, Server, Datastore) on it currently and we have m5.xL running image server. I did some testing on a development environment and down graded each running a m5.xL for the Enterprise Core stack and M5.L for the Image server. both struggled with performance. Maybe if you had a more distributed system you might be able to get away with it but would only perform for the lightest weight applications.