An Intel-VMware Perspective: Intelligent Queuing Technologies for Virtualization

Table of Contents

Bigger Burdens for Hypervisors ...................................2

Queueing Technology Overview ...................................3

Receiving Packets ......................................................3

Transmitting Packets ..................................................3

VMDq Performance Use Case Scenario .......................4

Summary ..................................................................4

How to Get VMDq .......................................................4

How to Get VMware NetQueue ....................................4

Bigger Burdens for Hypervisors

Deploying virtualized environments on more powerful platforms is

a growing practice among IT departments in order to consolidate

server workloads and reduce data center footprints. However, this

practice can have a significant impact on system and application

performance as workloads increasingly depend on network I/O.

While IT managers are adding greater processing power and

reducing the infrastructure footprint, this kind of consolidation

does not necessarily mean more efficient network throughput

in the virtual environment. A balance between system perfor-

mance and networking capabilities is required to achieve

optimal application services from consolidation.

In virtual environments today, the hypervisor manages network

I/O activities. With more virtual machines (VMs) and increased

traffic through the platform, the hypervisor requires more CPU

cycles to sort data packets and route them to the correct VM

(Figure 1), reducing CPU capacity available for applications.

Intel’s VMDq is a breakthrough technology that reduces

the burden on the hypervisor while improving network I/O

performance through the virtualized platform.

Figure 1. In virtual environments today, the hypervisor

manages network I/O.

Hypervisor Software

Server Hardware

SW Switch

NIC

vNIC vNIC vNIC

LAN