The practice of cloud rendering in urban IoT projects

In the early stage, we deployed rendering directly on the server, but high-performance terminals were needed to support rendering tasks, and the GPU performance of the client terminal was very high. Customers were limited by the convenience and cost-effectiveness of accessing terminals, and would not purchase high-configuration computers in batches just for display needs. Accessing 3D scenes is not convenient, and applications need to be repeatedly deployed on the demonstration machine during presentations, which is inconvenient to share work results with others. The early deployment method could not meet the actual needs of customers, so we used cloud rendering technology in the urban IoT project.

Part 01

The concept of cloud rendering

Cloud rendering technology was first proposed by AMD. Similar to cloud computing, it places the 3D rendering process, which has high performance overhead, on a remote server. Users issue operation commands through the client to access 3D resources. The server performs corresponding operations based on the user's commands and transmits the rendering results back to the user's terminal for display in real time.

Part 02

Advantages of Cloud Rendering  

Cloud rendering technology is used to place rendering tasks on high-performance servers, so users can access high-quality 3D images smoothly through web pages. At the same time, cloud rendering can provide users with the ability to access the project large screen on different terminals, and is perfectly compatible with computers, mobile terminals and other devices, helping users to view the large screen anytime and anywhere through various devices, making it more convenient to control project progress and facilitate communication between all parties on the project on the same large screen.

Part 03

  Difficulties in implementing cloud rendering  

1️⃣ The cloud rendering system consumes a lot of hardware performance when performing 3D rendering. Therefore, when facing multiple user requests, the server is required to have extremely high graphics processing performance, otherwise it will not be able to meet the simultaneous rendering requests of a large number of users.

2️⃣ The 3D scene needs to respond quickly to the commands input by the user. If the time from the user inputting the command to the corresponding action on the screen is too long, it will cause a noticeable sense of delay. The command delay depends on the network access performance of the user terminal and the server.

3️⃣ 3D rendering generally requires users to feel smoother at above 30fps. The higher the video resolution, the clearer the video, and the larger the amount of data transmitted. To achieve a resolution of 1920×1080, the transmission bit rate must generally be stably higher than 10Mbps.

Part 04

Production practice of cloud rendering  

At present, the OneNET urban IoT platform adopts a private deployment method for most projects in the field of urban IoT, which means that the large screen of the project will not have a high concurrency demand. The performance of high-end GPUs currently on the market can meet the deployment and delivery requirements of cloud rendering projects; in terms of network, according to data released by the China Communications Standards Association, in the fourth quarter of 2021, the Internet experience rate of users of fixed broadband in my country reached 62.55Mbps, and this network access speed can fully meet the requirements of cloud rendering. In production practice, we use UE4-based pixel streaming technology, which can more easily implement the technical implementation of cloud rendering.

Part 05

Project results of cloud rendering  

With the help of cloud rendering technology, we successfully loaded and displayed 30G of factory oblique photography data in the browser. Customers can smoothly access the 3D scene through the computer browser without installing any plug-ins, and integrate 2D chart data to display factory demand information. At the same time, combined with the factory's focus, the scene can be scaled, rotated, moved, and other operations, successfully completing the delivery of the smart factory project.