Optimizing Data Center I/O Virtualization
Executive Summary
Rapid adoption of virtualization technology in the x86 server market is bringing a wider array of applications into the virtualized IT infrastructure. Driving this adoption is the need to fully utilize computers, manage those resources more efficiently, and reduce the number of server "footprints" in the data center. All of this brings business benefits, such as reduced power/cooling requirements, reduced IT staff requirements, and greater IT flexibility. Computing tasks can be provisioned to available resources, as needed — and resources can be put to other uses, as business requirements change over time.
The rate at which data can flow from one device to another — long referred to as I/O, for input/output — is fast becoming a bottleneck in this growing virtualized infrastructure. If I/O is not sufficient, then it could limit all the gains brought about by the virtualization process.
With powerful new server hardware and software optimizations in the hypervisor, virtual machine (VM) density is increasing and is consolidating more I/O traffic onto the server bus. Also, many users are beginning to virtualize high-performance tier 1 applications that can be I/O intensive. High-speed I/O devices, such as solid state disks (SSDs) and 10Gb Ethernet (10GbE) links, will bring new levels of I/O performance but will further pressure existing I/O limitations. The evolution of converged fabrics in the datacenter will increasingly merge cluster, storage, and network traffic on a single physical fabric, requiring a high-performance I/O framework to support high throughput levels. This I/O fabric will also underpin the new internal cloud data center architecture that is an emerging, and maturing, model for "private clouds" within the enterprise.
Read the full Optimizing Data Center I/O Virtualization Case Study.
140KB
很抱歉，此 PDF 仅供下载
Optimizing Data Center I/O Virtualization
Executive Summary
Rapid adoption of virtualization technology in the x86 server market is bringing a wider array of applications into the virtualized IT infrastructure. Driving this adoption is the need to fully utilize computers, manage those resources more efficiently, and reduce the number of server "footprints" in the data center. All of this brings business benefits, such as reduced power/cooling requirements, reduced IT staff requirements, and greater IT flexibility. Computing tasks can be provisioned to available resources, as needed — and resources can be put to other uses, as business requirements change over time.
The rate at which data can flow from one device to another — long referred to as I/O, for input/output — is fast becoming a bottleneck in this growing virtualized infrastructure. If I/O is not sufficient, then it could limit all the gains brought about by the virtualization process.
With powerful new server hardware and software optimizations in the hypervisor, virtual machine (VM) density is increasing and is consolidating more I/O traffic onto the server bus. Also, many users are beginning to virtualize high-performance tier 1 applications that can be I/O intensive. High-speed I/O devices, such as solid state disks (SSDs) and 10Gb Ethernet (10GbE) links, will bring new levels of I/O performance but will further pressure existing I/O limitations. The evolution of converged fabrics in the datacenter will increasingly merge cluster, storage, and network traffic on a single physical fabric, requiring a high-performance I/O framework to support high throughput levels. This I/O fabric will also underpin the new internal cloud data center architecture that is an emerging, and maturing, model for "private clouds" within the enterprise.
Read the full Optimizing Data Center I/O Virtualization Case Study.
