Sponsored By

The Fall of Fibre ChannelThe Fall of Fibre Channel

The Fall of Fibre Channel – Stephen Saunders

November 6, 2001

18 Min Read
The Fall of Fibre Channel

The Roman Empire.

The Dallas Cowboys.

Van Halen...

Once upon a time, Great Institutions. Now… Not!

This just in: Fibre Channel is set to join them in the “Where are they now?” file.

I’ll repeat that for the hearing impaired: Fibre Channel will fall. Fail. Falter. In other words: The end is nigh!

But all kidding aside, I’m not trying to start a panic here. As is the case with our Sun (that hot, throbbing blob up in the sky; not the one that makes servers), there’s still some time (tick, tock, tick, tock) before the thing fizzles out. Network managers who’ve deployed Fibre Channel, or investors that have bought stock in Brocade, still have time to prepare for the end of the world (as we know it, Jimmy!).

But how long? Read on, gentle reader, and all shall be revealed.

Let’s get one thing clear from the start: I’m not anti-Fibre Channel. Farfrom it: I was actually one of its earliest proponents. Way back in 1989 (simpler, yet happier, times),just before I moved to the U.S.A. from the U.K., I wrote one of the firstarticles about what was then a little known high-speed LAN standard calledFibre Channel.

Fibre Channel was the brainchild of a bunch of academic types at the American National Standards Institute (ANSI) withvery large foreheads (“fiveheads,” you might say), and its biggest competitor backthen was generally considered to be asynchronous transfer mode, or ATM.

Anyway, 12 years ago, it was a younger, thinner Steve Saunders that did thedue diligence for his article, compared the specs and applications for bothstandards, and stuck his cervical vertebrae out – predicting that Fibre Channel would beatout ATM for very-high-throughput I/O apps within ultra-high-speed LANs (theterm “SAN” not having been invented at that time).

I even titled the article: “Fibre Channel Outpaces ATM.”

And, my! How the villagers laughed. Suffice it to say that my analysis wasunpopular with vendors, pundits, and other journalists – most of whom atthat time were hailing ATM as the Second Coming.

Whatever… It took a decade, but both the ANSI boffins and I were eventuallyproved right, which is quite satisfying if you’re the sort of “told you so”person who takes satisfaction in that sort of thing (and I know I am).

But now things are about to change, and Fibre Channel’s fortunes are set totake a nasty turn for the worst. How do I know? It’s all about history – andlearning from its lessons.

Flashback time

As a journalist I’ve written about a bunch of technologyrumbles over the years: shared media networks versus switched networks,100Base-T versus 100VG-AnyLAN, ATM versus IP, Netware versus NT (not much of a fight, that one).

And, of course, Ethernet versus Token Ring, FDDI, FDDI II, FFOL, HIPPI,IsoEnet, ATM, and a few others that should never have gotten off thewhiteboard.

(For more on all this, see if you can lay your hands on a copy of thatripping historical yarn, The McGraw-Hill High-Speed LAN Handbook, by StephenSaunders, McGraw-Hill, 1996, available at all good garage sales andrecycling facilities near you.)

What’s interesting (I hope) about all these fights is the characteristics they share:

  • One:

    • The press picks one technology and starts hyping the bejeesus out of it. Thepack’s choice is based almost exclusively on what veeps of marketing tellit, though trade journalists also have a natural tendency to migrate towardsthe more recent of any two given technologies. This is because the “NewTechnology Underdog” story is a lot more interesting than the “Hey, This NewTechnology Sucks!” story, and actually takes a lot less chutzpah to write. (Note: The press can’t lose with this gambit, because, should their chosentechnology keel over later on, they still get to write the juicy “Backlash!”“Where Did It All Go Wrong?” and “We Blame [name of big vendor]!” articles).

    • Two:

      • Vendors line up on either side of the battlefield,William-Wallace-versus-Evil-English fashion, and…

      • Three:

        • If there’s a standards process involved, the vendors promptly hijack it,putting their short-term fiscal interests over those of their users.Meanwhile…

        • Four:

          • Market research firms forecast that the market for this new technology willgrow to twenty gazillion-trillion-jillion dollars by next Tuesday, and…

          • Five:

            • End users (poor buggers) pay way too much to wander around trade showsand conferences in a confused state, not being invited to the standardsmeetings, trying to work out which is the best option, and finding out that, yes, actually, sometimes people do get fired for buying[insert name of current or historical market-leading vendor here: IBM/Cisco/Novell/Brocade, etc.).

            So what’s missing from this five-step program? Ironically, it’s the onething that can genuinely provide a clue as to which way the tech tides areset to flow: an examination of the business case – now, and in the future,for the various technology options.

            Next: Case, business, lack ofAnd that brings us to Fibre Channel. Specifically, the problem with Fibre Channel is that, while it is both reliable and fast, it doesn’t actually present users with a
            business case at all!

            When it comes to data centers, it’s simply the only game in town – a technology hegemony, essentially. And that makes it vulnerable to new technologies that can demonstrate the all-important ROI (return on investment).

            What’s aggravating matters is that a single vendor dominates the FibreChannel market: Brocade Communications Systems Inc. (Nasdaq: BRCD) accounts for about 80 percent of the marketfor Fibre Channel switches of eight to 16 ports, or about 60 percent of theSAN equipment market total.

            As we all know (another of history’s lessons) hegemonist dictatorships don’t have a greatrecord of putting their citizens’ (read: end-users’) interests first. Whenever one vendor is runningthe show – be it IBM Corp. (NYSE: IBM) with mainframes, Cisco Systems Inc. (Nasdaq: CSCO) with routers, or Brocade withFibre Channel switches – two things are inevitable: One, prices go up (andstay there). Two, proprietary gimmicks rule the day, making interoperabilitya concern, and making it harder for other vendors to get into the game. The inevitable end result is customer dissatisfaction and an open door for newer, cheaper, disruptive technology options.

            Come the Revolution

            Unless you’ve been living in a cave, Gollum-style, for the last few months,you’ll know that the technology that’s emerged to challenge Brocade’s FibreChannel dominance is IP/Ethernet storage networking.

            This isn’t a fat pipe dream. Scale Eight Inc., for example, is already making anice nut by offering file software to companies, including both Microsoft Corp. (Nasdaq: MSFT)and Viacom Inc. (NYSE: VIA), that want to set up rich media services over long-distance IPnets. It’s signed three new customers in the last three weeks – and claimsits revenues in its last quarter increased 80 percent over the preceding one (see Scale Eight Set to Close $30M).

            Things are happening on the hardware side as well. Cisco, Nishan Systems Inc., Pirus Networks, SAN Valley Systems Inc., 3ware Inc., and others are working on IP storage switches (see Nishan Swaps Out Sales Team). Quantum|ATL Products Inc. recently demo’d iSCSI running on its tape drives. Meanwhile, disk drivemanufacturers like EMC Corp. (NYSE: EMC) and Hitachi Data Systems (HDS) plan to start building iSCSI (SCSI over IP) interfaces intotheir equipment once the spec has been implemented in silicon – notsoftware. That day isn’t far off. QLogic Corp. (Nasdaq: QLGC) says it will ship all-ASIC iSCSIimplementations in the first half of 2002. In the meantime, Adaptec Inc. (Nasdaq: ADPT) and Alacritech Inc., amongstothers, are now shipping pre-standard SCSI adapters that use a hybridASIC/CPU design to provide decent performance, while allowing theflexibility to upgrade the final standard when it’s ratified (see Show Time for iSCSI).

            Of course, simple availability of products won’t create a market on its own.What’s needed is that all-important business case. And that’s exactly why Ibelieve IP/Ethernet technology will be successful in its challenge, thoughnot for a while.

            Time to get specific. While IP technologies have absolutely no chance ofdisplacing Fibre Channel in the data center today, there are two otherplaces where they will make immediate inroads:

            • First, for use in building distributed storage networks, or backing up databetween remote locations. Compared to the two other options companies nowhave for this application, the business case for IP in this topology is acomplete no-brainer. Installing an IP storage switch to shoot data over anISP network can be done for a tiny fraction of the amount it takes to do thesame thing by purchasing Fibre Channel equipment and leasing dark fiber. It’s also an order of magnitude faster than the other option: using CTAM (Chevytruck access method).

              • The second place where IP storage will take off is in smaller sites andserver farms that could use a bit of a performance push, but where it wouldbe way too expensive to deploy Fibre Channel equipment. In these environs,IP-based technology will provide a perfect “SANs for the rest of us”solution, allowing smaller companies to enjoy the sort of performancebenefits now only enjoyed by Fortune’s Gang of 100.

              In other words, IP is set to “top and tail” the SAN market, making small buttactical inroads into Fibre Channel territory.

              These small victories are important because they will provide a technologyapprenticeship for IP storage – during which time vendors can iron outproduct problems, standards committees clear up the FCIP, SOIP, iSCSI, iFCPmess, and network managers get some familiarity with the technology (see SNIA's FCIP Group Debuts).

              How long will this period last? Two to three years. That’s when anotherevent will kick IP storage into overdrive: the general availability of 10-Gbit/s Ethernet products. These devices will for the first time provide theinfrastructure muscle for IP to take on Fibre Channel in the meat of themarket: the data center (see Gigabit Ethernet: Hope vs. Hype).

              Remember: IP storage technology won’t be new then. Network managers willhave used it, and learned to trust it. This means that when 10-Gbit/sarrives Fibre Channel vendors will have to compete against it without helpfrom their best friend, Mr. FUD (fear, uncertainty, and doubt). Worse, theywill be up against vendors offering all of the generic benefits of IP andEthernet technology (interoperability, low cost, ease of use,manageability). Which of course, they can’t do, because their Fibre Channelproducts will still be proprietary, expensive, hard to use, and asmanageable as your average kindergarten class.

              For now, as we’ve noted, Fibre Channel enthusiasts still have some run time before them. The technology’s not imminently due to join Wayne & Schuster, Earth Shoes, and the Third International in the Dustbin of History. In fact, it’ll probably remain with us indefinitely in a gradually diminishing role. Think of the Royal Family. Or lava lamps...

              Where does all this leave the vendors? Obviously, IP storage represents ahuge opportunity for companies like Cisco and Intel Corp. (Nasdaq: INTC). Qlogic also stands todo well, if it can continue to steal away Brocade’s Fibre Channel marketshare while simultaneously selling arms to both sides (Fibre Channel and IP)in the technology war (see QLogic Sees Good Times).

              But overall, this technology transference will result in a lot more losersthan winners. While the long-term prospects for IP storage are good, aninflux of too much VC capital has meant that too many startups have startedpursuing this market too early. In fact, the shakeout has already started.

              (For more, tune into the continuing adventure series: Things Getting Grim at Entrada, Disaster for Entrada? , Entrada Faces Its Demons, and – who could forget? – Entrada's Last Gasp?)

              And then there is Brocade, for whom IP represents a “good news, bad news”kind of deal.

              First, the good news…OK. I lied – there is no good news for Brocade.

              So let’s turn to the bad news (always more fun, eh?), most notably the factthat Brocade is in a tough spot when it comes to adapting its strategy toaccommodate IP.

              Whatever should it do? One option would be to follow Cisco’s example and buyone or more startups in the new space. Has Brocade bought an aspiring iSCSIvendor? Not that I’ve noticed. Instead, it has chosen to partner with Ciscoon developing an IP blade that will sit in its Silkworm 12000 SAN platform, which – alas! – isn’tshipping.

              Now, there’s an analogy I once heard about putting foxes in charge of chicken coops… I don’t recall exactly how it ran, but, if memory serves, it’s considered a not very sound policy, all in all. It’s prettyobvious to me (and, I hope, to you) that partnering with a company that is setto become your number one competitor within the next three years for a keytechnology is… well, stupid, not to put too fine a point on it (see All Eyes on Cisco).

              [Note: If you’re so benighted as to disagree, you might perhaps be morecomfortable reading Storage Networking World, rather than Light Reading.]

              Stupid, Stupid, Stupid

              In a recent chat with Light Reading's sister publication Byte and Switch, Brocade’s VP of marketing, JayKidd, said that the 12000 will not ship with the Cisco blade, and that hedidn’t know when it would be available – noting that Cisco was in charge ofits development, not Brocade.

              As far as internal IP hardware developments go, Brocade doesn’t have any.Kidd says Brocade believes it can implement IP using merchant silicon, ifnecessary, and that it has no plans to develop an IP equivalent to its FibreChannel silicon.

              In other words, Brocade doesn’t have any immediate plans to doanything about IP. Judging from its strategy (to use that term in its broadest possible sense),the company isn’t too fussed about IP at all.

              Instead, it’s sticking to its Fibre Channel guns. In the short term, this iscertainly the easier option. Any overt support by Brocade for the IPparadigm (excuse me; I had Pompous Analyst for lunch) might shake itsshareholders’ faith in the One True Technology andBrocade’s divinely ordained rule as Lord of the Data Centers nowand evermore, Amen!


              The point is: In the long term, this is likely to turn out to be a very, very baddecision. Let’s go back to our history lessons.

              Brocade’s attitude smells strongly of Cabletron Systems Inc. (NYSE: CS) – which took too longto convert its product line from shared-media to switched technology – and the late, lamented Bay – which forgot to develop a chassis-based LAN hub, resulting in itgetting its clock cleaned by Cisco.

              But Brocade most recalls Madge – the token ringvendor, not the manicurist – which fought a war against Ethernet way back in the mid-90s. Rather thancover its technology bases, Madge chose to evangelize its own approach. Itlost: In the space of a year Madge went from being one of the most importantcompanies in networking to a footnote, while its founder, the eponymous (pardon the expression) Robert Madge, ended up quitting to quietly ride his horse 'round his farm in rural Blighty.

              All very well for our Bob, but a poor precedent, not to say paradigm, for any technology company.

              Let’s look a bit more closely at what’s happening at Brocade.

              Busy, Busy, Busy

              Brocade’s position on IP extends to saying publicly that it will representonly a “small fraction” of its business for the next few years.

              Ergo, it must continue to dominate the Fibre Channel market if it is tomaintain its revenues (and share price) and not go the way of companies likeEMC, Foundry Networks Inc. (Nasdaq: FDRY), or — Heaven forfend! – Lucent Technologies Inc. (NYSE: LU). The problem is that Brocade is already losingmarket share. And given the number of companies that are gunning for a pieceof its pie, things are set to get worse.

              Of course, Brocade could just tell people that this doesn’t matter becausethe Fibre Channel market is growing faster than it can lose it. But thissort of argument is unlikely to impress Wall Street.

              A better defense would be a strong technology offense, by which Brocadeaggressively develops new Fibre Channel products, brings them to marketfirst, proves that they are best in breed, and then forces the competitionout of business by using its strong cash position to undercut their prices(à la Cisco).

              So is Brocade doing this? Not exactly. Its foray into 2-Gbit/s Fibre Channelswitches is a case in point (see Brocade Unveils Fabric Switch).

              Consider the story so far:

              To its credit, Brocade has been investing heavily in research anddevelopment. In fact, over the past five quarters, Brocade's R&D-to-salesspending ratio (a good barometer of a how aggressively a company is developing new products) has averaged 19 percent – higher than that of othermarket-leading companies like Cisco (17.5%), QLogic (17.3%), EMC (11.9%), orNetwork Appliance Inc. (Nasdaq: NTAP) (12.6%).

              Unfortunately, this hasn’t resulted in Brocade winning the race to ship a 2-Gbit/s switch – both Vixel Corp. (Nasdaq: VIXL) and QLogic have beat it to the punch (see Vixel Ships 2-Gig Switches and QLogic Ships 2-Gig FC Switch). QLogicrepresents an especially big problem, because it OEMs its 2-gig hardware toother equipment manufacturers, spawning multiple new competitors forBrocade. McData Corp. (Nasdaq: MCDT) is also stepping up the pressure, making inroads with Compaq, one of Brocade's key OEMs (see Compaq Markets McData).

              Add to this laggard behavior the – admittedly, unconfirmed – rumors that Brocadehas run into problems getting the security and zoning codes to work on itsnew, larger 12000 switch, and if I were a Brocade shareholder I’d probablybe wanting to know where all that R&D money is going. (Perhaps Brocadehired someone from NASA or some other federal agency to run its R&Ddepartment. Who knows?)

              Brocade’s pricing strategy for its new products also leaves something to bedesired. Instead of choosing to cut its competitors’ strings by pricing itsproduct at the low end, it’s made the classic monopolist’s decision to tryto milk even more out of its market dominance. Sure, Brocade’s new 2-Gbit/sswitch only costs around 15 percent more than its 1-Gbit/s products, but it’s twice the price of some of its competitors’ 2-Gbit/s equipment.But the part of Brocade’s approach to the 2-Gbit/s market that really gets mygoat (and I think we all know how painful that can be!) is its attitude to product testing.

              As it happens, Byte and Switch has spent the last six monthsorganizing the world’s first multivendor test of 2-Gbit/s equipment. Twovendors took the challenge, and their names shall be revealed when wepublish the test in December. Suffice it to say that Brocade is not one ofthem. And it’s come up with some priceless excuses for not takingpart (see Is Brocade's SilkWorm Losing the Thread? and Brocade Ponders 2-Gbit/s Test).

              To begin with, it told us that it couldn’t spare any of its new productsbecause they were all being shipped to customers. In the six year’s that Ihave been running test programs, I’ve found one thing to be absolutelycertain. When a vendor says, “Our entire inventory has been shipped tocustomers,” this may be translated in either of two ways: “We don’t havea product,” or “We don’t think our product has a chance in Hell to win your test.”

              (This is oneof the three “big rules” in life, the others being: “Never start a land war inAsia” and “Never buy a used Japanese car from a friend.”)

              Things got even better last month when Greg Reyes, Brocade’s ebullient CEO,told us Brocade was too busy to participate (see Greg Reyes, Chairman and CEO, Brocade). Oh, right,that’s OK then.

              (One wonders which of the great networking breakthroughsmight never happened had their inventors decided they were “too busy.” When Alexander Graham Bell called, “Watson, I need you!” suppose for a moment that Watson had been “too busy”... filing his nails, perhaps, or – who knows? – having it off with the chambermaid. Why, we’d all be talking into soup cans today!)

              Now, it’s fine to turn down an invitation if you’re the small, shy, shrinking violettype of vendor, one that doesn’t go around climbing on every available soapboxtelling the world that you are the Greatest, the Best, the Number One, theChamp. But Brocade, and more specifically, Reyes, does just this (see Brocade CEO in Tick-Top Form).

              And to give credit where its due, nobody does it better. Reyes has done aterrific job rallying his staff around the Fibre Channel flag, andproselytizing it to shareholders in entertaining public appearances. Suchrousing rhetoric is all very well, but if, when someone offers you theopportunity to back up your words with proof, you don’t take it, thenyou’re not a leader – you’re a blowhard.

              Reyes used to work for Banyan [ed.note: who?]. While he was swinging on hisVines, first Novell and then Micro$oft stole the company’s market from underit.

              Has he learned his lesson? History will judge.

              — Stephen Saunders, Founding Editor, Light Reading

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like