Cisco Blogs


Cisco Blog > Data Center and Cloud

One Big Happy Fabric

Was recently reading this article about FibreChannel over Ethernet, something we have written just a few times about :) I came across this article today on the technology and how it is coming along in the standards bodies and with a full slate of multi-vendor support coming in behind it.The really interesting part of FCoE is going to be how the adoption/evolution works. The way we are looking at it is fundamentally how do we evolve our existing director line to support FCoE interconnects with our Nexus line. The Nexus will primarily be interconnecting hosts/initiators, while the MDS will primarily be interconnecting arrays/targets. I think this is a rational adoption and evolution plan because at least historically IT departments seem to adopt new technologies on hosts more quickly than on targets. Additionally there is the outlier of a groundswell effect…If PCIe converged network adapters are available this year, they will be used primarily for test and qualification of an FCoE infrastructure. Test the Nexus interconnecting with hosts, JBODs, then into SANs via a separate VSAN, and then into wholesale production. The groundswell happens in 2009 and onward. When the 10GbE FCoE technology is integrated into the server mainboard with LAN on Mainboard ports and new servers start shipping with this technology available, by default the game will change.Fundamentally every server will have a Virtualized I/O. Similar to what the VMWare Hypervisor did to abstract O/S from Server Hardware allowing any workload to be located on any server; our Unified Fabric virtualizes the I/O so that any server can be software provisioned to have the I/O characteristics it needs to process the workload you would choose to locate on that host.Hypervisors enabled software provisioning of server CPUsUnified Fabric enables software provisioning of Server I/Odg

Comments Are Closed