Client GPU Redirection for Lower-Cost Virtual Desktop AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost of providing GPU capability in virtual desktop environments due to the expensive nature of GPUs and the inefficiency of deploying GPU arrays in data centers, leading to increased pricing for users.
Innovation Solution
Redirecting GPU capability from client computing devices to virtual desktops by utilizing the local GPUs already present in client devices, which are often idle during remote processing, and offloading machine learning workloads to these GPUs through a GPU controller and consumer system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If GPU arrays are deployed in data centers to provide GPU capability to virtual desktops, then GPU functionality is available to users, but the cost increases significantly
Solution Approach 1:
Instead of deploying GPUs in data centers to serve virtual desktops, the patent inverts the approach by utilizing GPUs already present in client devices. The GPU capability is redirected from the client device to the virtual desktop environment, eliminating the need for expensive data center GPU arrays while maintaining full GPU functionality for machine learning workloads
Solution Approach 2:
The patent makes client device GPUs serve dual purposes: local applications and remote virtual desktop workloads. By enabling the same GPU to handle both local and remote processing tasks through capability redirection, the system eliminates the need for separate GPU infrastructure in data centers, thereby reducing costs while maintaining versatility
2Productivity
If data center GPU arrays are deployed to support machine learning workloads, then processing capability is improved, but the complexity of infrastructure increases
Solution Approach 1:
The patent extracts the GPU processing capability requirement from the data center infrastructure and relocates it to the client device. By taking out the GPU function from the centralized infrastructure and placing it at the edge (client device), the system maintains high processing capability for machine learning workloads while dramatically simplifying the data center infrastructure
Solution Approach 2:
Instead of physically deploying GPU arrays in data centers, the patent creates a virtual copy of the GPU capability through software-based redirection. The GPU controller on the client device and GPU consumer in the virtual desktop environment work together to replicate GPU functionality over the network, eliminating the need for physical GPU hardware in the data center while maintaining processing capability
Data Source
AI summary
Techniques are described for redirecting GPU capability from a client device to the virtual desktop. The virtual desktop client includes a GPU controller deployed on the client computing device, which is capable of virtualizing the local GPU of the client device and exposing it to the virtual desktop. The virtual desktop agent operating on the host server includes a GPU consumer, which is capable of accepting machine learning (ML) or artificial intelligence (AI) workloads on the virtual desktop and offload these workloads to the GPU controller on the client computing device. When the GPU consumer detects the ML workload task on the virtual desktop, it transmits the ML workload task over the network to the GPU controller on the client computing device, which processes the ML workload task using the GPU of the client computing device and sends the results of the processing to the GPU consumer.