Petascale Cloud Supercomputing for Terapixel Visualization of a Digital Twin

摘要: Background: Photo-realistic terapixel visualization is computationally intensive and to date there have been no such visualizations of urban digital twins, the few that exist looked towards space rather than earth. Objective: our aims are: creating a scalable cloud supercomputer software architecture for visualization; photo-realistic 3D IoT data supporting daily updates; rigorous evaluation supercomputing application. Method: we migrated Blender Cycles path tracer public within new framework designed scale petaFLOP performance. Results: demonstrate can compute in under one hour, system scaling at 98% efficiency use 1024 GPU nodes delivering 14 petaFLOPS. The resulting image supports interactive browsing city its wide range sensing scales. Conclusion: resource available greater anything on national supercomputers providing access globally competitive resources. direct financial cost access, compared procuring running these systems, was low. indirect cost, overcoming teething issues with development, should reduce significantly over time.