This paper presents mechanisms and optimizations to reduce the overhead of network interface virtualization when using the driver domain I/O virtualization model. The driver domain model provides benefits such as support for legacy device drivers and fault isolation. However, the processing overheads incurred in the driver domain to achieve these benefits limit overall I/O performance. This paper demonstrates the effectiveness of two approaches to reduce driver domain overheads. First, Xen is modified to support multi-queue network interfaces to eliminate the software overheads of packet demultiplexing and copying. Second, a grant reuse mechanism is developed to reduce memory protection overheads. These mechanisms shift the bottleneck from the driver domain to the guest domains, improving scalability and enabling significantly higher data rates. This paper also presents and evaluates a series of optimizations that substantially reduce the I/O virtualization overheads in the guest domain. In combination, these mechanisms and optimizations increase the maximum throughput achieved by guest domains from 2.9 Gb/s to full 10 Gigabit Ethernet link rates.

Virtualization has fundamentally changed the data center network. The last hop of the network is no longer handled by a physical network switch, but rather is typically performed in software inside the server to switch among virtual machines hosted by that server.In this paper, we present the concept of a sNICh, which is a combination of a network interface card and switching accelerator for modern virtualized servers. The sNICh architecture exploits the proximity of the switching accelerator to the server by carefully dividing the network switching tasks between them. This division enables the sNICh to address the resource intensiveness of exclusively softwarebased approaches and the scalability limits of exclusively hardware-based approaches. Essentially, the sNICh hardware performs basic flow-based switching and the sNICh software handles flow setup based on packet filtering rules. The sNICh also minimizes I/O bus bandwidth utilization by transferring, whenever possible, inter-virtual machine traffic within the main memory.We also present a preliminary evaluation of this architecture using software emulation. We compare the performance of the sNICh with two existing software solutions in Xen, the Linux bridge and Open vSwitch. Our results show that the sNICh outperforms both of these existing solutions and also exhibits better scalability.