Hitesh Ballani, Dinan Gunawardena, and Thomas Karagiannis
The shared nature of the internal network in cloud datacenters is at odds with their multi-tenant nature. It leads to variable network and application performance, and allows for malicious behavior, both of which are key barriers to cloud adoption. To address this, we argue that the network should satisfy two complementary goals-- provide minimum bandwidth guarantees to tenants while bounding the maximum impact they can have on the network. This ensures the network is shared in an effifficient and robust fashion. This paper shows how careful VM placement and weighted bandwidth allocation can be combined to achieve these goals, and balance the competing requirements of tenants and providers.
Through large-scale simulations and a testbed implementation we show that minimum bandwidth guarantees, apart from helping tenants achieve predictable performance, also improve overall datacenter throughput. Further, bounding a tenant's maximum impact can curb malicious behavior.