vPipe: One Pipe to Connect Them All!

Many enterprises use the cloud to host applications such as web services, big data analytics and storage. One common characteristic among these applications is that, they involve significant I/O activities, moving data from a source to a sink, often without even any intermediate processing. However, cloud environments tend to be virtualized in nature with tenants obtaining virtual machines (VMs) that often share CPU. Virtualization introduces a significant overhead for I/O activity as data needs to be moved across several protection boundaries. CPU sharing introduces further delays into the overall I/O processing data flow. In this paper, we propose a simple abstraction called vPipe to mitigate these problems. vPipe introduces a simple “pipe” that can connect data sources and sinks, which can be either files or TCP sockets, at the virtual machine monitor (VMM) layer. Shortcutting the I/O at the VMM layer achieves significant CPU savings and avoids scheduling latencies that degrade I/O throughput. Our evaluation of vPipe prototype on Xen shows that vPipe can improve file transfer throughput significantly while reducing overall CPU utilization.