SHARE

Virtually Speaking: Unclogging the I/O Pipes

Let’s face it, there’s nothing terribly scintillating about I/O. At best, it’s akin to air: When data is flowing, it’s something no one thinks about. But when data flow slows, all eyes focus on the bottleneck, and I/O congestion is a frequent culprit. At present, I/O is perhaps one of the biggest technological hurdles that […]

Written By

Amy Newman

Apr 10, 2008

Datamation content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More

Let’s face it, there’s nothing terribly scintillating about I/O. At best, it’s akin to air: When data is flowing, it’s something no one thinks about. But when data flow slows, all eyes focus on the bottleneck, and I/O congestion is a frequent culprit.

At present, I/O is perhaps one of the biggest technological hurdles that virtualization technologies face. With the capability to do more on the same box comes the inherent downside of overly taxed processors and with that the equivalent of rush-hour congestion or clogged pipes.

Not surprisingly, companies are springing up to address this demand. Some, like Xsigo and Egenera have built the solution into their hardware, bringing a proprietary air to the solution.

Although 3Leaf’s solution is currently an appliance, it uses standard hardware, making it easily modifiable.

“Our real value proposition and differentiator is that we use standard commodity hardware and can therefore follow technology curves in multiple areas, for example, processors and switch fabric,” Rob Reiner, senior director of marketing at 3Leaf told ServerWatch.

Last week, the company announced it had obtained a license for Intel’s QuickPath Interconnect, enabling it to build virtualization support for Intel servers. With AMD’s HyperTransport license already on board, 3Leaf now supports the full monty of x86 options.

This Monday, the company unveiled its larger vision: The Virtual Compute Environment (VCE), a fully virtualized data center infrastructure designed to deliver flexibility and mainframe-class availability to x86 servers while optimizing data center resources in terms of both efficiency and costs. When the VCE arrives, it will offer complete virtualization of server resources, easy repurposing of workloads, and fast and flexible provisioning.

When fully up and running, VCE will virtualize the entire IT infrastructure from CPU to memory to I/O, thus enabling server resources to be virtualized on demand and moved in a just-in-time fashion to and from applications.

Reiner outlined the two phases of VCE. The V-8000 Virtual I/O Server, which is available now is part of the first phase. The focus of Phase 1 is on the virtual I/O Server. A number of Fortune 100 companies and top-tier OEMs are evaluating the solution presently, Reiner said.

A key selling point of the V-8000 is its simplicity. The supported servers have no spinning media, no HBA/NICs, fewer drivers and types of drivers, 80 percent fewer SAN and networking ports, 80 percent fewer standby servers, and full redundancy along the entire data path.

Expandable servers fall under the second phase. A CPU/Memory Server is scheduled to arrive in 1H09, Reiner said. This second phase builds on many of the advantages that the V-8000 delivers and uses a commodity switch fabric.

Reiner described VCE as “disruptive.” Whether or not it really will be, the company is on to something with its recognitions importance of I/O.

As virtualization moves into the real world, I/O has the potential to be the fly in the ointment. Performance matters. No matter how well a solution virtualizes, if what is being virtualized cannot perform up to snuff, the virtualization solution itself is irrelevant.

Amy Newman is the managing editor of ServerWatch. She has been following the virtualization space since 2001.

This article was first published on ServerWatch.com.