Generally, I/O on virtual machines hosted in cloud services like AWS and Azure do not perform as fast as direct attached or SAN/NAS storage. The I/O operations themselves on cloud storage suffer from higher latency than properly tuned local storage. An observer of the VM I/O can see I/O rates drop to zero, even during periods of load being imposed by MongoDB to it's file system, a situation not expected for local storage. Given that we cannot affect how cloud storage is implemented by cloud providers, we should optimize our I/O for these cloud environments.