I've read a few articles recently announcing the impending death of iSCSI-based SANs everywhere. Over the past few years, iSCSI has gained major traction and attention as organizations look for ways to lower costs or gain the benefits of robust shared storage technology. Now, companies are announcing a new wave of Fibre Channel products based on a new standard called Fibre Channel over Ethernet (FCoE) that, like iSCSI, rides on Ethernet, thus removing one of the major hurdles that keep organizations from adopting Fibre Channel: complexity.
While iSCSI is a fantastic solution for many organizations, when it comes to sheer throughout and performance, the technology does not match what can be accomplished with today's Fibre Channel products. iSCSI is completely reliant on TCP/IP to handle transport needs, which introduces processing overhead into the storage equation. For environments in which throughput and low latency are absolutely critical — high-transaction environments, for example — iSCSI is probably not the best choice. However, I remain convinced, based both on my own experience and from my reading, that iSCSI is a great choice for a great many other kinds of environments.
FCoE aims to break the iSCSI adoption deluge by simplifying the infrastructure. iSCSI's claim to fame is twofold: 1) It generally costs less, and 2) it costs less because it relies on existing protocols and infrastructure for the brunt of its work. With Ethernet as the underlying transport and TCP/IP used as the communications mechanism, iSCSI brings storage to the masses as most organizations have someone in house with requisite skills.
With an FCoE infrastructure in place, native Fibre Channel protocols will ride on an Ethernet-based transport mechanism, thus achieving speeds of, today, up to 10 Gbps. Further, FCoE doesn't use TCP/IP; the protocol continues to use native Fibre Channel to communicate, although some relatively minor tweaks have been made in order for Fibre Channel to support Ethernet. The advantage in this lies in the much lower encapsulation overhead; wrapping SCSI packets into a TCP/IP construct in order to crate iSCSI can be a processor-expensive task. However, iSCSI's TCP/IP-based operation also gives it one key wide-area advantage: easy routing that won't be available in FCoE without special hardware, thus keeping FCoE inside the data center walls for now.
When I use the word "Ethernet" in the paragraphs above, don't think that this is the same Ethernet you're using for client and server connectivity; FCoE will rely on what some are calling "Enhanced Ethernet" and others are calling the "Converged Enhanced Ethernet" or "Data Center Ethernet" standard. Simply put, Data Center Ethernet aims to bring together today's disparate data center cabling standards of Ethernet, Infiniband, and Fibre Channel. Of course, converging these infrastructures will limit Data Center Ethernet products to currently shipping Ethernet speeds; today, that's 10 Gbps, with speeds of 40 Gbps and 100 Mbps under development. I'll write more about Data Center Ethernet in another posting.
I'll admit it right up front... I'm a big fan of iSCSI. It's a protocol that has allowed the "little guy" to enjoy the benefits associated with block-level shared storage, and support for the technology has become almost ubiquitous. I don't see FCoE as the end for iSCSI by any stretch of the imagination; in fact, I think we'll see iSCSI products hitting the market that can take advantage of this Enhanced Ethernet being created to support FCoE. Regardless of what transport it rides on, Fibre Channel is still more complex than TCP/IP and requires a different skill set to manage. I can see FCoE being very popular in existing Fibre Channel shops that do want a bit more simplicity brought to the operation, and I can see it being successful in smaller organizations that have massive storage needs. However, FCoE doesn't mean the end of the road for iSCSI development. In fact, as I implied, I bet we'll see a whole host of new iSCSI gear designed to make native use of 10 Gbps Ethernet or Enhanced Ethernet that will give FCoE a major run for its money.
Since 1994, Scott Lowe has been providing technology solutions to a variety of organizations. After spending 10 years in multiple CIO roles, Scott is now an independent consultant, blogger, author, owner of The 1610 Group, and a Senior IT Executive with CampusWorks, Inc. Scott is available for consulting, writing, and speaking engagements and can be reached at firstname.lastname@example.org.