Real-time cloud envisions the migration of time-critical applications to shared cloud infrastructures in order to leverage elasticity, scalability, and operational flexibility. However, this vision conflicts with core cloud design principles: deep resource sharing, multi-tenancy, and increasingly complex virtualization stacks.
In this talk, we analyze the main threats to predictability and security in multi-tenant cloud systems that hinder the deployment of time-sinsitive applications, with particular attention to shared memory subsystems and virtualization boundaries.
We then focus on two representative challenges: memory bandwidth isolation and hypervisor attack surfaces, examining recent research efforts as well as their practical limitations.
We conclude by questioning whether real-time cloud is a technically grounded objective, or a business-driven narrative that underestimates the structural constraints of modern cloud architectures.