Q: How do FCoE and iSCSI relate to one another?
FCoE is a low-overhead, high-performance encapsulation of Fibre Channel frames for delivery via lossless Ethernet and is primarily focused on data-center environments. iSCSI is designed to run over existing networks and relies on TCP/IP to provide acknowledgments, flow control, and retransmission when needed. While both are designed to deliver SCSI commands over a network, FCoE maintains all of the Fibre Channel protocols and structures to simplify the integration of FCoE and Fibre Channel into a common data-center environment. Because iSCSI uses different protocols and data structures, integrating iSCSI into a common data-center environment requires the use of a gateway to perform the necessary protocol conversions. Fundamentally, FCoE is Fibre Channel— with a different physical interface.
Q: Will FCoE eliminate the need for iSCSI?
Probably not. iSCSI has the ability to run over existing Ethernet networks, as well as company intranets for long-distance applications such as remote copy or mirroring. FCoE is based on lossless Ethernet and will most likely be deployed over a new generation of products designed for the converged Ethernet environment. These products will support Ethernet enhancements such as Per-Priority Pause flow control, Enhanced Transmission Selection Quality of Service through bandwidth management and traffic prioritization, and Congestion Notification and Management.
Q: If iSCSI and FCoE co-exist, will this encourage the development of converged NICs that have TCP/IP offload engines for iSCSI and FCoE for FCoE?
We are already seeing more NICs that offload various functions. Once you get to 10Gbps, you don’t want the system’s processor trying to handle low-level link operations along with its normal functions. Whether functions are offloaded via dedicated hardware or a dedicated processor or core, this will certainly be the norm at 10Gbps.
While FCoE does not require a TCP/IP off-load engine, or TOE, it does benefit from off-loading certain Fibre Channel and SCSI protocol functions. With the increasing density of chips, I expect to see more functions offloaded from the system’s processor to the adapter.
Q: Will FCoE encourage a migration from Fibre Channel SAN networking to Ethernet?
I think we will see an evolutionary transition from today’s predominantly Fibre Channel SANs to more mixed environments in the future. As companies begin FCoE deployment, FCoE and Fibre Channel will co-exist for some time. If FCoE is as successful as it shows signs of being, I would not be surprised to see the ratio of FCoE links versus native Fibre Channel links swing toward a higher percentage of FCoE links. Most data-center managers are fairly conservative, so I expect this to be a gradual transition that follows the normal procurement and replacement cycles.
Q: Will FCoE provide an opportunity for Ethernet to become the storage networking platform at some point, for both block- and file-storage access?
Many years ago, I was at a technology conference and one of the presenters stated that Ethernet would ultimately win out over all other networking technologies, including Fibre Channel and InfiniBand. At the time, I thought he was being overly enthusiastic, but if we look at what has been happening with things such as Voice over IP and the transfer of other types of information over Ethernet, it may be that he was more prescient than I gave him credit for at the time.
Q: Will the present SAN management superstructure on SAN directors from vendors such as Brocade and Cisco be replicated on Ethernet switches?
Given that the FCoE Forwarders are basically Fibre Channel switches with Ethernet ports, the SAN management structure will be present there. I don’t expect that enhanced Ethernet switches will adopt the Fibre Channel management structure.
Q: Does FCoE give an entry into storage networking for Ethernet switch vendors not currently active in storage networking?
The challenges for Ethernet switch vendors will be both technical and operational. On the technical side, there will be (enhanced) Ethernet switches that snoop the FCoE protocols to automatically configure Access Control Lists and Static Forwarding Table entries to provide a more robust storage environment. There will also be FCoE Forwarders that are essentially Fibre Channel switches with one or more Ethernet ports (and maybe one or more native Fibre Channel ports).
The easiest opportunity for existing Ethernet switch vendors will obviously be through enhancements to existing switches to support protocol snooping. Providing the full FCoE Forwarder function requires providing all of the Fibre Channel protocols, functions, and services, such as the Name Server, Login Server, Fibre Channel address assignment, Zoning, FSPF, etc. Providing the latter functions essentially involves creating a Fibre Channel switch with Ethernet ports.
On the operational side, most storage solutions are sold as just that—complete solutions. It is very difficult to sell a point product into the data-center environment. Getting a new product certified and supported by key storage providers, and accepted by the ultimate customer, could present a challenge.
Q: Will some kind of virtualized FCoE NIC be produced so that multiple VMs in a virtualized server can have logically separate communications through it?
We already have this today for both Ethernet and Fibre Channel adapters, and it is reasonable to expect that the same capabilities will exist in Converged Network Adapters for FCoE.
Q: How do you rate the various FCoE products?
I think it is still too early to rate FCoE products. The InterNational Committee for Information Technology Standards [INCITS] T11 committee is still in the process of formalizing the FCoE standard, so the current implementations are all essentially “pre-standard.”
Existing Fibre Channel providers, such as Emulex and QLogic, certainly have a substantial base of Fibre Channel experience to draw upon, but they may find the Ethernet world challenging. Vendors with an Ethernet background will need to develop Fibre Channel competency to support FCoE. I think that both camps will face some interesting challenges and opportunities as they enter this new converged networking world.