My preference is for the group to concentrate on network rather than systems virtualization, but I also definitely agree with Thomas. On Mon, Oct 12, 2009 at 8:15 AM, James Kempf <james.kempf at ericsson.com> wrote: > Another topic is "over the top" v.s. "close to the metal" network virtualization. Talking to GENI researchers, they seem to be planning to run virtualized networks as overlays like X-Bone was. This is in contrast to the, shall we say, "original promise" of virtualized networks, which is that they provide isolation between traffic on different "slices". Providing isolation while still providing the cost benefits of statistical multiplexing, especially across service providers, is a challenging research problem. Most L2 VPNs with service guarantees are expensive and time consuming to provision and confined to a single provider, which is probably why academic researchers aren't planning on using them. And statistical multiplexing is not possible because the VPN customers ping constantly to make sure they are getting their SLAs fufilled. Fwiw, I'm working in GENI with my FlowVisor virtualization layer project (tech report newly available at http://www.openflowswitch.org/wk/index.php/FlowVisor if anyone is interested), and while I'm not sure where you draw the line between "over the top" vs. "close to the metal", FlowVisor (intuitively to me at least) would seem to run close to the metal and is not really an overlay like X-Bone. Cheers, - Rob .
Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.