Verizon recently has been testing new GPU-based technologies which will enable just one graphics processor in offering services to several clients. The carrier believes that these technologies will benefit the VR, XR and AR users and also AI/ML users, gamers and real-time organizations.
Overall, these capabilities may pave the path for one new category of budget-friendly mobile cloud-based services and help development of low cost, scalable edge-based GPU compute.
Verizon’s 5G technology and its edge network are basically designed to offer cloud services at the network’s edge closest to the user. Due to the heavy graphics and imaging that may benefit using this new technology, many applications will be able to run much better on one GPU. AI/ML, VR, AR, XR, real-time organizations, and AAA gaming are very much dependent on the GPUs for their commute capabilities. However, the scarce availability of proper resource management with regards to GPUs poses a barrier to deployment of these technologies.
Thus, to meet this requirement, Verizon developed one prototype using GPU-slicing techs and virtualization management which supports any kind of GPU-based cloud service and allows multiple-client loads as well as tenants.
Verizon tested this new GPU-orchestration prototype on one live network located in Texas and found that it allowed for 8 times more concurrent users. In fact, the orchestration allowed 80 times more users when utilized for one graphics gaming service.