I really enjoyed reading Marc Farley’s, blog on FCoE. I couldn’t agree more with many of his very passionate observations.
But I wanted to add my thoughts. First, I DO believe that folks will comingle storage (iSCSI or FCoE) traffic and standard IP networking through the same physical link. They will NOT however, commingle that traffic on the same logical link. Customers will utilize 802.1q VLANs to segment traffic as recommended today and into the future. Customers will also use 802.1p priority tagging to ensure that storage traffic receives priority over other traffic. Ten gigabit Ethernet changes everything and delivers on the promise of a unified fabric, and the customers I have talked with are itching to remove the 4-6+ gigabit Ethernet and 2-3 fibre channel ports, the associated tangle of cables with two 10GbE ports, two cables, and only two switch ports. Regardless of protocol, that where the money savings are for customers, both in terms of component and power savings, labor, hassle, and support costs. BTW, you don’t have to wait for some newfangled protocol to achieve this today…there is this great solution that can make this happen today…iSCSI!!!
Second, I think the underlying Ethernet technology that powers FCoE, Datacenter Ethernet (DCE)*, is a rising tide that will float all boats. It is REQUIRED for the primitive FCoE protocol, but will benefit iSCSI and any other protocol that leverages Ethernet, like iSCSI. Provided you want to spend the money on new infrastructure, DCE will definitely improve the Quality of Service (QoS) characteristics of Ethernet.
Last, FCoE will be valuable for customers with a large existing investment in fibre channel. These customers will benefit from the cost and complexity reductions associated with a unified fabric. They can and should adopt iSCSI for new storage infrastructure build outs, particularly for storage servicing their increasingly virtualized compute farms. Finally, they can utilize FCoE to facilitate access to their legacy FC investments until they are fully depreciated and retired. Because of that, you will see Dell shipping DCE capable NICs/HBAs/Switches, FCoE forwarder systems, and more.
Per my previous posts, we at Dell believe that iSCSI is a superior storage fabric and there are tons of reasons why. The future arrival of DCE and FCoE does not change that, but we believe that FCoE is valuable for our many customers with legacy FC environments and provides a path to a fully unified fabric for those customers.
Last note, the savings associated with a unified Ethernet fabric and 10GbE are real. Why wait for FCoE, you can achieve those savings TODAY with iSCSI?