Wrong Time for an End Run
When you buy "standards-compliant" storage products—before standards are finalized—you run the risk of significant upgrade costs later.
Two years ago, while meandering through 115 cable channels on a Sunday afternoon in search of something worth watching, my wife Margaret and I came across an NFL football game. I'd quit watching football after the 1987 strike, blaming everyone from team owners to their gazillion-dollar players for interrupting my weekly amusement.
But on this Sunday, my wife said that although she was interested in the game, no one had ever explained it to her. We watched the game; she asked questions; and I answered. That's all it took to addict her—and me—to what has become a shared Sunday (and Monday night) ritual.
It took a while, but I think I understand her fascination. Margaret has discovered the "end run"—a football play for surmounting an obstinate defensive line by running the ball around it to one side or the other. In her case, football provided a tactical end run around many things—including my occasional preoccupation with work that can distract me from hearth, home and her.
End runs can be a good thing. Sometimes they can get you around some significant problems and take you where you want to go. But the end run can also be a questionable strategy, especially when it comes to storage technology. Today, an alarming number of vendors are engaging in end runs around the technology standards-setting process at the American National Standards Institute (ANSI) and the Internet Engineering Task Force (IETF) by releasing products that are "100 percent compliant" with standards that do not yet exist.
| Making a Storage Purchase? 4 Steps to Take Now You're probably going to run into challenges confirming vendor claims about standards compliance. Here are a few tips: - Get to know what standards are in development at the Internet Engineering Task Force and ANSI—the brokers of all standards development efforts related to TCP/IP, Fibre Channel and SCSI. Both organizations have Web sites—IETF's IP Storage Working Group maintains its documents at www.ece.cmu.edu/~bassoon/ips/Docs/docs.html; ANSI's T-11 Committee is at www.t11.org. The IETF IPS Working Group also maintains an e-mail list that you can join for free to monitor the traffic between engineers working on iSCSI, iFCP and FCIP standards. If you want to subscribe, send an e-mail to: ips-request@ece.cmu.edu. Be aware that you may receive a lot of e-mail at times.
- Check with a trusted reseller or integrator who handles the products of numerous competitors. Chat with fellow IT practitioners at professional conferences and forums.
- Talk to third-party vendors, especially storage management software vendors whose products must work with a broad range of hardware.
- Read the trade press publications.
—J.W.T. |
Standards Still Developing
IBM Corp., Cisco Systems Inc., Emulex Corp. and a number of other vendors already have announced products to support iSCSI, a SAN protocol still under development at IETF's IP Storage Working Group. The vendors were busily demonstrating interoperability at last spring's Storage Networking World conference and will no doubt show off shipping products during fall trade shows like Networld+Interop and Comdex.
Truth be told, however, none of these products is truly "iSCSI-compliant" because, at present, there is no iSCSI standard to comply with. At the time of this writing, the draft specification was moving into its seventh iteration. From the flak traffic on the IPS reflector, an e-mail list for the IP Storage Working Group, there are still a formidable number of nontrivial technical issues to surmount before iSCSI will become a recommended standard.
The IPS WG has been working toward a September deadline for draft RFC submission of its three IP SAN protocol specifications-in-development. The three specifications are Fibre Channel over IP (FCIP), Fibre Channel over Internet (iFCP) and SCSI over IP (iSCSI), according to co-chairman David Black (who also serves as a senior engineer for EMC Corp.). However, few members of the IPS Working Group believe that all issues will be worked out to meet this deadline.
The same holds true for the ANSI T-11 Committee's development of the new Fibre Channel Switch Interoperability protocol standard, nicknamed FC-SW 2. In February, there was a burst of announcements from vendors that said they were already releasing "FC-SW 2-compliant" switches—or soon would be. But, these announced products also comply with a standard that is not yet a standard.
What the vendors were really announcing, according to one insider, was that the switches were implementing the "T-11 letter-ballot approved version" of FC fabric log-in routines and the "universal version" of Fibre Channel Shortest Path First (FSPF) routing algorithm—both of which are components of the draft standard that are "unlikely to change radically" between now and the formal adoption of the standard at ANSI. FC switch vendors are doing their best to hype the gains that these provisions contribute to the problem of switch interoperability.
By no means, however, is FC-SW 2 a done deal. The standard is also supposed to include provisions for a common World Wide Name Zoning implementation that vendors can leverage to ensure that zones set up in SAN fabrics established with their switches will "merge" with zoning information created on other switches. Interoperable zoning is just as important as common fabric log-in and common FSPF to interoperable FC fabric switching.
Like the IETF's IP SAN protocols, release of the final FC-SW 2 standard has been delayed. It was also due to move up for public comment and standard adoption in September, but will likely take several months longer. You should also clearly understand that FC-SW 2 does nothing to guarantee FC SAN product interoperability with switches that are not standards-compliant, or with host bus adapters, controllers or other SAN devices that are consistently a source of expense, hassle and delay in FC SAN deployments.
| New Storage Standards Here are some of the components that the ANSI T-11 Committee's new Fibre Channel Switch Interoperability protocol standard (FC-SW 2) draft standard contains, which vendors are prematurely rushing to meet: - Fibre Channel fabric log-in routines
- A universal version of the Fibre Channel Shortest Path First (FSPF) routing algorithm
- Provisions for a common World Wide Name Zoning implementation.
—J.W.T. |
Making an End Run
As a byproduct of its SAN virtualization strategy, DataCore Software delivers an end run around many FC SAN interoperability problems, according to Nik Simpson, product marketing manager for the Ft. Lauderdale, Fla. company.
According to Simpson, FC SAN interoperability requires initialization and registration of node devices connected into the fabric, the sharing of zoning information and routing information, and the normalization of error codes broadcast by FC devices themselves. Many FC switches, as well as HBAs and controllers, do not perform these tasks in the same way.
He also notes that even if a product meets standards, that doesn't ensure common implementations by vendors or product interoperability.
DataCore's virtualization strategy entails the use of "Storage Domain Servers"—commercial server platforms dedicated to the virtualization and allocation of disk storage to SAN-attached hosts. DataCore's SANsymphony software "runs as a network storage control layer on top of the domain server's native operating system." SANsymphony, according to Simpson, handles the interoperation of devices as part of its virtualization method.
He says that the company had to come up with its own strategies for dealing with the differences between FC products in order to make its storage domain server independent of SAN devices. DataCore's customers can leverage the work that the company has already done to reduce SAN interoperability issues.
Vendors claim that the standards process sometimes takes too long and that customer requirements can't wait. However, there are risks for consumers who buy standards-compliant products when standards have not been finalized. One compelling risk is the increased likelihood of a requirement to perform a significant upgrade—sooner rather than later—in order to glean the benefits of a truly standards-based storage infrastructure.
End runs, as my resident football expert will say, nearly always run the risk of a behind-the-lines tackle and loss of yardage. And, of course, there's always the risk of dropping the ball altogether.