Global Robot Deployment


#1

Hi - is anybody currently using robots in global locations accessing an Orchestrator that is only deployed in one location? For instance, the Orchestrator is setup in the US, but robots are connected to the Orchestrator from UK, China, India, etc. Are there any latency or performance issues with this setup, or do all robots process similarly?

Currently our robots, Orchestrator and Elastic Search servers are all setup in the same location, and the applications that they access are local as well. We are looking to access applications in other locations and allow more users access, but are wary of latency and performance. I’m interested to know if anybody is using UiPath this way, or if you setup additional infrastructure for each location.

Thanks
Dan


#2

Hi,

From my understanding…

When you set up your Robots in Orchestrator you set them up to run your process on a particular machine. The entire process is running on that machine so the performance is limited by the performance of that machine alone. However, if you are accessing network shared locations then that could slow some things down. But, if you are just questioning whether the process slows down, it will run directly on that machine so you shouldn’t see any performance issues related to the location of the machine.

Lastly, you will find some performance of loading Orchestrator over the network and possibly Logs updating, but like I said, the process performance runs on the machine directly so you should be ok there.

Thanks.


#3

Thanks - good points. I would expect that a robot with a large amount of logs may see some degradation.

Is anybody running robots live where the robot and orchestrator are not in close proximity to each other?


#4

We are still relatively early in our process, so no measured performance (and no benchmark). But running Orchestrator in internal data centers and robots in Azure. There is some lag on startup, but run seems just fine. We are actually doing a couple of API calls in a custom extension as well as using the log and still seems pretty snappy.

Are you seeing something that is giving you pause? Or preemptively looking?


#5

Thanks for the info - just looking preemptively at this point. We are still fairly early in our process as well, but trying to provide an idea of what to expect prior to testing.


#6

Standard network issues will arise sooner or later - especially connection drops are a pain to deal with, as you might end up with transactions being stuck InProgress, or receiving “weird” errors. This is not dependant on being far away, it’s just that there’s more things in the middle that can drop the connection.

Robot work in itself shouldn’t be impacted for the most part, as it locally works between UiRobot and UiRobotSvc, which then handles communication with Orchestrator.
What you might want to be wary of is asset fetching, which by definition needs to wait for the response to continue. Same with queue operations (add/get/set/update status).

Do remember that if you’re crossing borders, data protection laws might apply and for certain kind of data it’s not allowed to send them abroad (where abroad might be defined as f.e. “outside EU”), at which point it might be required to set up a local Orchestrator anyway.