Virtual chip architectures, virtual I/O -- what's the limit?

Charles Babcock, Editor at Large, Cloud

September 14, 2007

2 Min Read

VMware seized the lead in virtualization by generating software that mimics the x86 instruction set, the internal instructions that operate Intel and AMD chips. But there are virtualization advocates who say much could be gained by virtualizing assets outside the x86 base.

One company, Transitive, virtualizes the instruction sets of many chips, including IBM's Power and Sun's UltraSparc. Transitive's QuickTransit virtualization software lets operating systems and their applications run on hardware for which they were never intended. But there's no Star Trek-style universal translator: Each QuickTransit product translates from one specific operating system and chip combination to another--say, from Solaris/Sparc to Linux/Intel Xeon, or vice versa.

InformationWeek Reports

Transitive CEO Bob Wiederhold says that, for several years, QuickTransit has translated Adobe Acrobat running under Windows into instructions meant not for the Intel x86 instruction set as Adobe originally intended, but for older Apple Macintoshes. Until recently, Macs ran on the IBM Power chip and used its instruction set. QuickTransit did the translation from Power to x86 effectively enough that most Macintosh users didn't realize they were running an alien, virtualized chip architecture.

Now, with new Macs running Intel processors, the need is reversed. "We're behind the new Intel-based Macintoshes running applications that were designed for Power-based Macintoshes," he notes.

The 7-year-old company has enjoyed significant behind-the-scenes success. A year ago, IBM agreed to offer a version of QuickTransit to its System p server users to let them run Linux x86 applications on IBM Unix servers.

VIRTUAL I/O
Few have conceived of input/output as needing virtualization, but piling up virtual machines on multicore servers strains those servers' capacity to deliver I/O. Xsigo Systems, a well-funded startup, thinks it can solve that and squeeze more performance from virtualized servers.

Normally, server I/O traffic is collected on network interface cards and host bus adapter cards, each one for a specific purpose, such as an Ethernet NIC or a Fibre Channel storage NIC. Each is limited in bandwidth.

Xsigo I/O Director supplies custom versions of NIC and HBA cards (at $8,000 a pop) in the server that virtualize the cards' traffic. Instead of routing the traffic to its intended destination, the custom cards route it into a $30,000 hardware box, where the Director software recognizes the type and amount of traffic coming in and can allocate switching resources to route it toward its proper destination. The server thinks it's talking to a NIC, but the NICs and HBAs have been virtualized, with Director software taking over and managing the virtualized card traffic. Xsigo claims Director can multiply a server's I/O capacity by 100.

While x86 server consolidation is a major step toward virtualizing the data center, even bigger gains in server use may still await virtualizing assets beyond the x86 box.

About the Author(s)

Charles Babcock

Editor at Large, Cloud

Charles Babcock is an editor-at-large for InformationWeek and author of Management Strategies for the Cloud Revolution, a McGraw-Hill book. He is the former editor-in-chief of Digital News, former software editor of Computerworld and former technology editor of Interactive Week. He is a graduate of Syracuse University where he obtained a bachelor's degree in journalism. He joined the publication in 2003.

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like


More Insights