Chapter 4. Sanity Check on IP SANs

In the last chapter, we examined the Fibre Channel fabric and compared it to the ideal SAN originally described in the Compaq Computer Corporation's 1997 Enterprise Network Storage Architecture (ENSA) white paper, which provided the original definition of a storage utility infrastructure. The chapter surveyed the myths that were created around "Fibre Channel SAN" topology by a cadre of overzealous vendors and concluded that, while the Fibre Channel protocol could be used to create a serial interconnect fabric that could solve certain problems, FC SANs delivered none of the business value associated with ENSA-type SANs.

It was a classic case of oversell: Vendors correctly perceived how the SAN concept resonated with prospective customers and pressed the Fibre Channel Protocol, the fastest available interconnect for open systems storage at the time, into service as the plumbing for SANs. Despite the fact that the vendors themselves knew the protocol was unsuitable for building real networks and came up short in nearly every service category required for true cost-savings, FCP ”and products built on it ”provided the means to sever dependency on the parallel SCSI bus, enabled the attachment of more storage devices, and placed devices at greater distances from the server host.

More importantly, the "new" protocol and topology enabled a group of vendors that formerly specialized in selling mainframe channel extension products and ESCON director switches, and whose fortunes (like those of mainframe vendors generally ) were hard hit by Y2K jitters and the flight from mainframe hosting that occurred at the end of the millennium , to breathe new life into their companies. To this day, most Fibre Channel switch vendors were yesterday 's channel extension product vendors.

The truth is that vendors knew from the outset that the Fibre Channel protocol was not up to the task of building real ENSA-style SANs. Asking their representatives point-blank questions regarding the erroneous business value claims they were making about FC SANs usually produced wry grins and acknowledgement of the lack of important network services in FCP. The universal response was less apologetic or defensive than simple and direct: "Yes, we know all that, but we're the only game in town."

For a time, the mantra of the Fibre Channel Industry Association seemed to be that any technology, regardless of how half-baked, was ready for business "prime time" if businesses would buy it. Burgeoning data, plus a "bull market economy," created a need for more storage and a willingness among consumers to invest in just about any technology that was presented to them and contextualized as "the next big thing."

Vendors advised the critics of FC SANs (including this author) that they were lagging behind the trends. Supply-side industry analysts validated vendor claims of the solvency of FC SAN value, despite the fact that most had never seen a real FC SAN, and contributed to the not-so-subtle bending of technical terminology to mean anything a vendor decided it meant . In short order, the Open Systems Interconnect (OSI) definition of a network was cast to the wind, as was the ENSA description of a storage area network, as vendors applied the term to describe basically any interconnect topology, including the configuration of disk inside an array cabinet! A SAN became whatever a vendor said it was ”a "marketecture" term with little or no technical meaning or relevance.

Just as Fibre Channel advocates were corrupting the meaning of SAN, a working group was being established within the Internet Engineering Task Force (IETF) tasked to develop a set of protocols for harnessing TCP/IP networks to become the plumbing for storage networks. The IP Storage Working Group was quick to dismiss arguments by the Fibre Channel Industry Association that IP could never serve as a storage interconnect. Protocol-specific issues, while technically nontrivial, could be surmounted, the group argued, and storage could be operated successfully as an application across an IP network.

By 2003, vendors in the storage industry were beginning to rally behind the newly completed Request for Comment (RFC) ”IETF lingo for a standard ”for SCSI over IP (iSCSI) and referring to FC fabrics in casual conversation and conference presentations as "legacy SANs." Some were the very same vendors who had just completed deployments of "legacy" Fibre Channel SANs for their customers!

Like FC SANs, current IP-based SANs also fail to deliver on the ENSA SAN vision. However, IP SANs move closer to the goal by utilizing a full-service network with a robust set of services as the plumbing for storage I/O.