Setting up everything on premise - Requirements

[T0Bi]

Hey there,

I’d like to set up nearly the whole UiPath Suite on premise, Orchestrator, AiFabric, OCR, Data Manager, etc.

As servers are not that cheap, especially when powerful ones are needed (for AIFabric), I was trying to find out exactly what I need in order to not overpay.

This has been a major hassle.

Yes, there is documentation about hard- and software requirements of the different parts, but it’s hard to combine all the information available, especially when it’s not 100% clear which parts you can run on the same server and which you can not.

Imagine I want to have Orchestrator + everything for Document Understanding (AIFabric, DataManager, OCR), what’s the cheapest way to deploy this on premise for testing purposes (not production)?

Did I get this overview right?

Could you please validate this and/or share your own experience with setting up UiPath on premise?

Cheers,
T0Bi

[Maneesha_de_silva]

Hi @T0Bi ,
Let me help you with this.

Please note that I’m writing under the condition you said as "on-premise for testing purposes ".

yes, I feel what you have mentioned, and how important your servers and utilization, this cannot be assigned as mention because of its consume a lot of hardware resources.

what you need to do is,

• You can couple App & DB Server in a single server (you can set up both app and DB {4cores and 8GB ram more than enough} )

• When considering (CPU-6 RAM-24 DISK-200) generally enough, but consider how much records you are going to do on testing, so then (even less cpu 4 , ram 12/16GB ram is enough for to run)
  And they strongly recommend installing AI Fabric on VM instances provided by popular cloud providers such as AWS, Azure, and GCP. you have to consider that also.and ubuntu is ok but I prefer windows if possible as my personal view that will minimize your other issue and compatibility complexity.

“There is no universal value for how much resources one ML Skill/Pipeline job will consume”
This 52 RAM base on as below
|||
|—|—|—|
|Core Services|10|
|Minimum for Serving (ML SKill)|2|
|Minimum for Training (Pipeline)|4|
|DU model Serving|4|
|DU model Training (1000 images)|24|
|GPU|8|

And for this another thing i suggest you is , you can try out the enterprise trial of Uipath AI Fabric on Cloud

All the hardware will be depend of the number of robots and the traction , if you trying to do less robot and less transactions for testing you can reduce the hardware assign , only the effect is processing time will bit of increase

I hope you are not worrying about the performance and processing time on this .

For OCR Engine and Data Manager is not a big deal , so you have several options

1. Put OCR Engine and Data Manager in a Single Server (Core-1 Ram-2 enough)
   2.Put OCR Engine and Data Manager in other server as you created (DB Server/AI Fabric)
   3 Create Separate 2 servers for OCR Engine and Data Manager ((Core-1 Ram-2 enough) enough for each)

And almost i suggest , start using Cloud servers & enterprise trial they offer , that will be a big solution to save our hardware and when you go for the production of there should be similar hardware requirement has to be their
and will give you a secret tip
Never assign entire the resource as that mention (assign 80% mentioned RAM specially will be mostly enough and according to your volume and the performance level you can increase the hardware as you want(hope you are configuring on VMs))

Hope you got some points.

[T0Bi]

Hi @Maneesha_de_silva

your answer helps me a lot!

I think I’m going to start with as little different servers as possible and scale up if needed.

Usually we use the cloud version for everything, but with the whole GDPR stuff in Europe, many customers do not want to have their (and their customers) data handled by a third party, especially when you can’t control where the servers are.

That’s why we have to make sure our team is able to set up everything on premise as well.

Thanks for your help.

Cheers,
T0Bi

[Maneesha_de_silva]

@T0Bi
yes got your point and actually agreed with you.
Further, I can tell and suggest that, Bring the entire infra at the same time make much hardware consume and you may have to lot of cost for hardware and consider with the cost of the application…
I don’t know whether how about your management react for it .
So you better go for with part by part like

As the First phase, you can buy Orchestrator and Robots, and start your project
And as a second phase, you can think about purchase the AI Fabric model and other stuff.

and one another tip… if you working with OCR/ICR check the product of ABBYY {https://www.abbyy.com/}
and this also tightly couple with UIpath