Network reliability assurances across geographically distributed cloud resources

Develop abstractions and mechanisms that provide end‑to‑end reliability assurances—including latency and bandwidth guarantees—for virtual clusters spanning geographically distributed resources over the public Internet in cloud data center networks.

Background

The paper explains that modern cloud SLAs mostly cover compute and storage, while there is no broadly accepted abstraction or method that guarantees network performance isolation (e.g., latency and bandwidth guarantees) across cloud data center networks.

It stresses that providing reliability assurances is especially challenging when resources are distributed and traffic traverses the public Internet.