vSnoop: Improving TCP Throughput in Virtualized Environments via Acknowledgement Offload

Abstract

Virtual machine (VM) consolidation has become a common practice in clouds, Grids, and datacenters. While this practice leads to higher CPU utilization, we observe its negative impact on the TCP throughput of the consolidated VMs: As more VMs share the same core/CPU, the CPU scheduling latency for each VM increases significantly. Such increase leads to slower progress of TCP transmissions to the VMs. To address this problem, we propose an approach called vSnoop, where the driver domain of a host acknowledges TCP packets on behalf of the guest VMs-whenever it is safe to do so. Our evaluation of a Xen-based prototype indicates that vSnoop constantly achieves TCP throughput improvement for VMs (of orders of magnitude in some scenarios). We further show that the higher TCP throughput leads to improvement in applicationlevel performance, via experiments with a two-tier online auction application and two suites of MPI benchmarks.

A deep investigation into network performance in virtual machine based cloud environments

“…Also, the effect of computational load inside a cloud virtual machine and its effect on network performance and stability has yet to be quantified. It has been shown, in the vSnoop developed in Purdue University, that the TCP performance with Xen virtulazation can be improved by offloading TCP acknowledgements to the driver domain [11]. We however show that not only does TCP performance degrade when the physical system is under heavy use but the packet drop rate can increase by up to 20%, indicating TCP acknowledgement offloading alone is not enough to fix the underlying performance issues.…”

Enhancing TCP throughput of highly available virtual machines via speculative communication

“…Another interesting observation is the relatively high deviation of the attainable bandwidth, especially for longer streams. We believe this is related to scheduling issues, similarly how related work describes it in case of TCP reception performance of virtualized environments [16].…”

“…Dong et al proposed efficient interrupt coalescing for network I/O virtualization and virtual receive side scaling to effectively leverage multi-core processors [8]. The closest work resembling ours is vSnoop [16]. vSnoop offloads TCP acknowledgment handling to the VMM so that TCP reception performance doesn't suffer due to scheduling delays of virtual machines.…”

TOSEC: A TCP/IP Offload based Virtual Network Security Framework in NFV Environment

“…The idea is also adopted in virtualization environments. [19] improves TCP throughput of virtual machines via TCP acknowledgement offload to domain 0, [20] presents the solution of offloading TCP/IP stack from guest OS to the hypervisor (vmkernel) thereby bypassing the guest OS kernel.…”