In-Depth

Fibre Channel's Dog Days of Summer

Debate heats up over Management Information Bases

• By Jon William Toigo
• 07/10/2003

The politics of network storage standards never cease to amaze me. Recently, the IETF’s IP Storage Working Group message reflector has been ablaze with debate over the issue of Fibre Channel Management Information Bases (MIBs).

The situation was prompted by the submittal in early June of a draft standard for consideration by IETF covering “MIBs Standardization for Fibre Channel.” Interestingly, the authors of the draft were from Cisco Systems and its spin-off, Fibre Channel switch maker Andiamo Systems. Their stated goal was to have the IP Storage Working Group create standards for enabling the Simple Network Management Protocol to manage Fibre Channel fabrics, which, as previously observed in this column, lack any sort of in-band management mechanism of their own. Behind the scenes, however, one had to wonder what Cisco was really after.

Most of the Working Group chat centered on why IETF should get into the business of supporting a non-IP networking protocol like FCP in the first place. ANSI’s T11 Committee was traditionally responsible for FCP, not IETF.

One commentator made the telling observation that much of the language in the proposed draft did not comply with Fibre Channel standards already approved by ANSI. For example, the draft authors sought to standardize a MIB to enable “VSAN awareness.” They noted that the "concept" of VSAN (a proprietary Virtual SAN approach advanced by Cisco and Andiamo) had been introduced to T11, but that considerable work remained to be done by T11 membership to reach any agreement that VSAN-like technology was actually required for Fibre Channel and related ANSI protocols.

To date, the commenter observed, the only information that had been presented to T11 was a Microsoft PowerPoint presentation on the VSAN concept—far from what was needed to ensure an agreed-upon definition, let alone an interoperable standards-based solution for FC SAN's. In light of this, he argued, any MIB definition by IETF around the technology would be very premature.

The same logic held true for other MIBs described in the draft covering “Fibre Elements and TE_Ports,” Virtual Fabrics, Fibre Shortest Path First (FSPF) “configuration,” and Zone Servers—none of which are defined in current ANSI standards covering Fibre Channel. Between the lines, the comments screamed out: Cisco Systems, having had little success in getting the ANSI T11 Committee to buy into its approach as an industry standard, is now trying to co-opt the IETF standards-making process to advance its Fibre Channel agenda.

Other comments pondered the obvious: If the Fibre Alliance MIBs currently used to connect SNMP services to FC SANs, were as effective as ANSI T11 and the Fibre Channel switch vendors have been claiming they were for the past several years, why was there a need for new MIBs at all?

One of the draft authors responded that T11 lacked MIB expertise and noted that a bad MIB, the Fibre Alliance MIB, had been widely implemented in the industry and that network management of Fibre Channel devices suffered as a result. Surprisingly, the ANSI T11.5 committee chair agreed that the Fibre Alliance MIB was not very good, but he added that no MIB has ever been approved as a standard by T11.

Still a third set of voices declared that the entire process was out of order. One strong voice in the working group declared, “It looks more and more like interested parties want us (IETF) to rubber-stamp their flavor of FC management. I do not think that this particular WG is qualified to handle FC related MIBs. As I said several time[s,] we are the IP storage WG. And FC belongs to T11.”

Finally, conciliatory voices offered that, if the current MIBs were screwed up, why not fix them within IETF, rather than passing the buck back and forth with ANSI?

The points and counterpoints rolled in like storm clouds for several days and the issue remained hot, sticky, and unsettled by the time this column went to press. What is obvious is that the Achilles Heel of Fibre Channel is its lack of in-band management services and its continuing need for a secondary IP network running SNMP to manage the fabric and its elements.

By not choosing to aid in FC MIB development, IETF may seal the fate of FC SAN, and pave the way for the substitution of IP-based protocols like iSCSI as the SAN interconnect of choice. Supporting the development of FC MIBs proposed by Cisco and Andiamo may lead to the same result, while promoting Cisco-style SANs now and in the future.

In the IETF, the latter option may be viewed as the greater sin. Allowing a company to standardize its own vision or technology is tantamount to abandoning the vendor-neutral open standards-making process.

Of course, the IP Storage Working Group will also be casting a vote by not acting on the MIBs as well—albeit a more defensible one within their charter. Clearly, the group will be demonstrating preference for its own creature, iSCSI, over ANSI’s Fibre Channel. It sort of makes you wonder why iSCSI needs its own industry association.