When I am trying to connect to nucleus from another pc on the same network, It is failing to connect to the server, knowing that both of them share the same network/IP. What my be the problem in my case?
Please follow the sharing instructions in our docs.
Usage â Omniverse Nucleus documentation (nvidia.com)
Also, please make sure when you share the server out, you are using the hostâs IP for the server path, not localhost as the server name.
If you are having issues after youâve followed these docs, please let us know.
1 Like
How can we share files between Omniverse Nucleus servers across the public Internet? Weâre having trouble setting up a server which has the host IP for the server path. We canât (so far) locate where to do this. :)
Context: We have set up Omniverse on an on-premises PC with an RTX 3060. This PC is on the Internet, via local ISPs (about 600 Mbit/s connectivity) with a variable IP. Create and View work, but of course theyâre pretty slow and thereâs not enough memory for larger models.
On AWS, we have set up an Omniverse virtual workstation, which we access through DCV Nice. It is running on a g5 instance and of course rendering is very quick. We have in theory shared out a server (using localhost, which is apparently a problem; we should use the virtual workstation IP instead, which we have, but are unsure how to do this).
On the on-premises PC, within Create, we âAdd New Connectionâ, type in the virtual workstation server name, or the IP address, which in either case is followed by âlog in via Web browserâ but the pop-up never appears, and then thereâs an error message saying unable to connect, check internet connectionâŠ
The objective is to move USD files on the on-premises PC (which are created in Nucleus via a link from SketchUp) to the virtual workstation on AWS, which has a fixed IP.
Same question, but we also have Omniverse Enterprise set up inside an AWS VPC, with a reverse proxy server. How do we connect to these machines and share files which are created on the on-premises PCs?
1 Like
Hi @nicholas.bedworth Was it difficult to setup Nuclues server on AWS? Does it require Enterprise version? I am trying to do it following the documentation but it seems very complicated, programming involved etcâŠ
Would you please send any resources that you guys used to setup the AWS server? thanks
Karol,
Basically itâs fairly straightforward, from an IT perspective, but yes, there are a lot of details. An IT engineer will definitely be required, but the right person can set it all up in a day. The Omniverse developer support team is quite active on their discord server, so please ask questions over there; they can get you oriented.
The NVIDIA âhow toâ documentation (Drew Como) is similar to but more accurate than the corresponding AWS blog post. Fundamentally you set up a c5 instance for the reverse proxy, and also a c5 (see next post from Drew) as the Nucleus server. One of the differences between Omniverse Workstation and Enterprise is the latter allows more collaborators, and is available on a paid subscription basis; 30 day trial.
Hope this helps a little bit!
Nick
1 Like
Thanks a lot Nicholas! Appreciate it!
Youâre very welcome! On UpWorks, we hired Sahil Dubey in Pune, with good results. He knows what to do, especially having just completed similar work for us.
Hi everyone!
Thereâs a little confusion here that Iâd like to be clear up. There are two versions of Omniverse:
- Omniverse Workstation (Individual)
- Omniverse Enterprise
https://docs.omniverse.nvidia.com/prod_install-guide/prod_install-guide/overview.html
However, that will also break down into two versions of Nucleus as well. Thereâs a local Nucleus install that is part of the Workstation product, but there is also an Enterprise Nucleus Server. The main differences are how itâs licensed, used, and the scalability options. But please keep in mind the Workstation product is not really a âserverâ per se - while it does allow another person to connect to it (for collaboration), itâs not designed or meant to be used outside of a local setting.
The Enterprise version which is based on Linux and runs within Docker containers, is truly the Enterprise product. This version also does not require a GPU as itâs not providing any type of visuals; itâs just the server behind the scenes storing your data and providing the conduit for collaboration.
Hereâs a link to a page that further explains the differences:
https://docs.omniverse.nvidia.com/prod_nucleus/prod_nucleus/arch_overview.html#nucleus-distributions
The documentation for setting up an Enterprise Nucleus Server on AWS is here:
https://docs.omniverse.nvidia.com/prod_nucleus/prod_nucleus/enterprise/cloud_aws_ec2.html
The documentation for setting up an Enterprise Nucleus Server on MS Azure is here:
https://docs.omniverse.nvidia.com/prod_nucleus/prod_nucleus/enterprise/cloud_nucleus_azure.html
The documentation for an Omniverse Virtual Workstation is here:
https://docs.omniverse.nvidia.com/prod_nucleus/prod_nucleus/enterprise/cloud_aws_vdi.html
If you have any additional questions, please continue to post in this thread and I will respond within 24 hours.
2 Likes
This was a really useful post, with a practical explanation of what the products are intended to do. Thanks!
So basically the Nucleus is, as you said, a kind of advanced file system and sharing framework. When one attaches to it with workstations, the GPUs on these machines do the rendering. If the machine doesnât have an adequate GPU, one attaches to the Nucleus with a virtual workstation, and on it one can specify the GPU performance desired.
Questions:
- Is there/could there be a virtual workstation for Azure?
- The Azure Nucleus Enterprise install appears to be a somewhat simpler process that for AWS. Could we set up the Nucleus on Azure and access from AWS-based virtual or on-premises workstations? Seems to me the answer would be yes.
- If we want to use Spot instances for the GPU on a virtual workstation, how could this be set up? Letâs say we have a very large model to render, or several of them, Spot instances will be economical.
- Finally, if we need really large rendering capabilities, how can we connect to them from a workstation?