isolation from other services executing on shared resources. Latencysensitive applications
executing in the cloud therefore have highly-variable response times, yet attributing the
additional latency caused by virtualization overheads on individual requests is an unsolved
problem.