Info:
NVIDIA? Iray? Server is a multi-user, distributed rendering solution that focuses on flexibility and ease of use without sacrificing performance. You connect to Iray Server by using an Iray-compatible application.

You can use Iray Server for two purposes:

Queue Rendering: To upload a scene from a client application to Iray Server for offline rendering.

Iray Server maintains a queue of rendering jobs. After a rendering job is uploaded to Iray Server it is added to the queue. All jobs in the queue are rendered, one after the other, without further involvement from the client application. A client application can, in fact, be disconnected from Iray Server after submitting jobs.

Rendering results are stored by Iray Server. You can retrieve these results from the file system or by using a web browser.

Streaming: To offload interactive rendering from the client application to Iray Server. The client application performs no rendering operations. All rendering is performed on the machine where Iray Server is running. Images are streamed back to the client application.

Offload rendering makes sense when Iray Server is running on more capable hardware or when you want for free up GPUs and CPUs on your local machine. The only differences you may notice when offloading rendering is potentially faster rendering and less resource usage locally. Any changes you make to a scene from the client application are immediately reflected in the images streamed back from Iray Server.

All asset management is automatically performed by the client application and Iray Server. You do not need to manually upload any scene assets such as textures, materials, light profiles and so on. This is true, even when assets on the client are changed.

Iray Server is designed for efficient performance across a range of client-server configurations:

When a client application is based in the same local area network (LAN) as the machine Iray Server is running on
When Iray Server is running on a machine in a different network, or even in a different country. Both the upload of the scene and the streaming of the rendered images are designed to minimize the usage of your Internet bandwidth.
To minimize network and Internet usage, the client application analyzes the active scene when it connects to Iray Server. The client application only sends parts of the scene and scene assets which have not been rendered previously on this particular Iray Server installation. Uploads are independent of any naming, scene usage, or instancing conventions that are used.

This approach has a number of important benefits for you:

It minimizes upload times.
It makes interactive usage quite fluid.
It liberates you from asset management.
It guarantees that what you render locally will match what you get from Iray Server.