Here is the section of the report on Brocade and McData from the Rule the World newsletter from the previous post
Rule The World Newsletter
Matt Tomkins
April 2001/Vol. I/No. I
matt@canada.com
<snip>
Brocade: The Weakest Gorilla?
Introduction
Although it’s not quite a Gorilla in the technical Geoffrey Moore sense of the word, Brocade does currently exhibit many of the attributes prized by The Gorilla Game investor. With a dominant share in a market with strong barriers to entry and high switching costs, it is often near the top of investors’ future Gorilla Candidates list. Brocade makes switches for Fibre Channel based Storage Area Networks (SANs) and, according to market research, currently controls over 90% of its market. Despite its apparently comfortable position at the top of a fast growing market segment, which is expected to grow to around $8 billion by 2004, there are many dangerous clouds forming on the horizon. From an impending clash of the giants within the fibre channel market itself to the emergence of industry wide standards and the eventual long-term disruptive threat posed by IP based storage networks, there is perhaps no other dominant Next Generation Networks vendor that is in a more precarious position than Brocade Communications.
The Big Mac Attack
The most immediate threat to Brocade will come from increased competition on its own turf from some formidable adversaries. Many investors seem to be under the false impression that Brocade has a lock on the Fibre Channel switching market. However, the recent announcement of Cisco's secretive SAN switching initiative lends further proof that Brocade's position within the industry could still be challenged. Although the release of Cisco's SAN switch is still about 18 months away, it does demonstrate that the market is far from conceding this lucrative niche to the company.
Despite the obvious concern surrounding any company that becomes a target of the networking king, Cisco is not even Brocade’s biggest problem. It would probably come as a surprise to those not directly following the storage industry to learn that a relatively small company recently spun off from EMC could pose a significant threat to such a dominant vendor. Well, it’s time that Brocade investors stood up and took notice of McData since it’s about to become the company’s biggest competitor. Conventional wisdom amongst investors seems to be that with Brocade's 90% share of the SAN switching market, it would be virtually impossible for any competitor to overcome such dominance (unless of course Juniper decided to get into Fibre Channel switches ;-) but that's unlikely to happen). Unfortunately, not only is this opinion wrong but the statement itself is false. Contrary to popular belief, Brocade dominates only a portion of the SAN switching opportunity, not the entire market. When all types of Fibre Channel switches and hubs are included, Brocade's market share falls dramatically from around 90% to 38% (numbers as of 3Q00). While the company remains the overall market leader, its position no longer looks quite as secure. When this wider view is taken, McData suddenly jumps into the picture, taking second place, rising from 0 to a 22% share. Not only that but McData is arguably better positioned for the future of the networked storage market than is Brocade.
The reason why McData seems to be off of most investors’ radar is because, until recently, it was not considered one of Brocade's competitors. There are currently three different segments of the SAN switching market: Workgroup, Department and Director. Brocade's dominant position stems from its 90% share of the combined workgroup and director market. However, the company currently has virtually no share of the high end director market where McData holds an even more impressive monopolistic market share of over 98% (some estimates have been as high as 99.8%). Needless to say, neither company is faced with much of a competitive threat at the moment. If only the world operated with fixed boundaries then both Gorilla's could live happily ever after in their respective niches. Unfortunately, as corporations begin to focus on end-to-end switching solutions, the lines in the sand will blur and a monolithic battle for SAN supremacy will ensue.
In order to understand why this trend will occur, one must first understand the current SAN architecture and why it will soon change. This report assumes that the reader has a basic understanding of how a SAN works and why they are advantageous to corporations. The arguments are of little consequence if one does not have this fundamental knowledge. Anyone who is unclear on this subject should read the Network Appliance report from last year’s “Ten Companies That Will Rule The World” report published at the now shut down soapbox.com. Although the discussion of SANs is extremely limited it does provide a basic overview. This section of the report can be accessed free of charge at the following link:
tsrec.com (See the second page)
At the low end of the SAN market, workgroup switches are the primary networking technology utilized (dominated by Brocade). This environment uses an arbitrated loop topology, meaning every device is located on the same data path and thus share the bandwidth. Unfortunately, as devices are added to the network, bandwidth constraints soon arise, limiting the effectiveness of the looped architecture to only the smallest SAN environments.
The next implementation up the SAN hierarchy use department switches (dominated by Brocade), the most popular of which is currently a 16 port design. The advantage of department switches is that they connect devices in a fabric rather than a loop. The benefit of a fabric is that it offers full bandwidth to all devices and allows for far greater scalability. Department switches currently represent the majority of the market and are the primary reason for Brocade's perceived position as the Gorilla of the Fibre Channel switch. While the advantages of this type of architecture over the old DAS (Direct Attached Storage) systems were enormous, it did not yet represent the ultimate synthesis or nirvana that some IT administrators had foretold. Despite the advances, the majority of SANs remain isolated islands of computing, unable to share resources across the entire enterprise. Department switches lack the redundancy and scalability necessary for the implementation of these enterprise-wide SANs. This is where the Director class switch enters the picture.
McData owns the Director Class switch market with an almost total share (although Inrange does have a product). These products operate in the SAN backbone and are used to connect isolated islands of storage across the enterprise and are the only class of products that meets the needs of this area of the network. The current workhorse of this market is the 32 port design, with larger 64 port configurations currently being rolled out. As the trend towards fully networked storage continues, the demand for Director class switches will likely skyrocket. Enterprise Storage group estimates that around 75% of SANs are currently deployed as islands, resulting in an enormous growth opportunity for connecting these islands together. In fact, while sales of Department and Workgroup switches were almost triple those of Directors in 1999, IDC expects Directors to represent the majority of the market by 2003, growing from a 31% share to 54%. While there is some dispute over the exact growth rate of the market, there does seem to be a consensus that it is shifting from the edge to the core of the network. Therefore, it would seem that although Brocade is currently sitting in the sweet spot of the market,
McData is comfortably positioned at where the market is headed. Sitting back and enjoying their hard earned success is not an option for either Brocade or McData. They must continue to innovate and meet current and future customer demands. As the demand for complete end-to-end SAN solutions increases, both of these companies will need to begin offering a complete suite of workgroup/department/director switches or risk losing share to competitors. Unfortunately, this means crossing over onto each other's turf and competing head to head, a trend that is in the midst of occurring as Brocade begins to move upstream into the Director class market and McData moves down into the workgroup and department opportunity. Although it is impossible to predict the outcome of this battle, from their current positions, it would appear that McData has an early advantage.
As the amount of data and complexity of SANs increases, the primary requirements of IT departments will evolve away from simple connectivity and towards scalability and reliability, trends that should prove to favour McData, as its ASICs and software are already targeted at providing the intelligence and capacity necessary to control large amounts of data in a complex environment. Brocade on the other hand will need to add this functionality to its products before it can offer a competitive solution. How well the company will be able to scale its switches has yet to be seen, something that could pose a future risk
McData is also spearheading the SAN interoperability initiative, which hopes to allow customers to deploy products from multiple vendors into a single network. McData should be one of the primary beneficiaries of these interoperability standards as it will allow the company to compete more easily in the department switch market. Much of the hair on Brocade's knuckles stems from its use of proprietary technology that protects a great deal of its installed base from fair competition. The lack of SAN standards results in the inability to combine switches from different vendors into a single storage network. As a consequence, in order to migrate to a different vendor’s solution, a customer would need to replace the entire network. These high switching costs have become a significant advantage for Brocade since it blocks other vendors from gaining a foothold in the company’s installed customer base. Once these proprietary boundaries are removed and completely open standards are established, it should be much easier for businesses to add McData's switches to their existing SANs. As for Brocade moving up into McData's stronghold, it would seem that the technological barriers to entry are higher in the director market and should thus provide more protection against competition.
A likely result of the impending clash will be both companies losing share in their own current niches and gaining a position in the others’. Fortunately, the market for SAN interconnects will be enormous and should have no trouble supporting a two horse race. In fact, these two players could one day form a duopoly of sorts in their dominance of the market. History has shown that while customers demand a second source for products, there is often little interest in additional suppliers as they tend to complicate matters. This occurs because of the fact that companies must ensure interoperability between competing products so that they can be deployed together in the network. Ensuring that two products interoperate is relatively simple. However, as the number of suppliers increases, the number of interoperability tests that must be undertaken increases exponentially (since each product must be tested with each other), soon resulting in a significant cost and time burden for the companies. Nowhere is this trend more obvious than in the network core, where Juniper and Cisco currently act as competitive wedges for each other, helping to keep others out of the market. If this event occurs quickly, it might be enough to prevent Cisco from gaining significant traction in this explosive market opportunity.
The IP Disruption
Even if Brocade prevails in the battle against McData and eventually Cisco, it must still contend with the disruptive potential of IP based storage area networks that threaten to make the entire industry obsolete over the long-term. Historically, NAS has used Ethernet technology while SANs relied on Fibre Channel (and SCSI to a lesser degree). Each of these technologies offers certain advantages over the other. Primarily, block level commands over Fibre Channel and SCSI were best for certain tasks while file level commands over IP/Ethernet were best for others. However, over the course of the next few years, the lines between the two will be blurred as both sides try to incorporate the advantages of the other into an optimal solution. The SAN camp is currently pushing two standards for merging their Fibre Channel/SCSI technology with IP. The first is Fibre Channel over IP, which is not meant to convert the entire network to IP but simply to expand the distance limitations of a SAN over an IP WAN. Although Fibre Channel traffic can travel much further than the 25 meter SCSI restriction, it is still limited to around 10 km. This has resulted in the still restricted SAN islands issue, which, although superior to simple direct attached storage, is still unable to fulfill the ultimate goal of a fully networked storage infrastructure. Fibre Channel over IP mitigates this limitation by integrating SAN islands into the IP backbone WAN. The technology accomplishes this task by encapsulating Fibre Channel frames into IP packets and mapping Fibre Channel fabric domains to IP addresses. The result is a network topology that maintains the existing Fibre Channel infrastructure but is able to convert frames to packets in order to transfer the information over the common WAN backbone to another SAN across city or across the world where it is re-converted into its original Fibre Channel frames. This development will allow users from any location to easily access SANs over the Internet. This technology is not so much a significant threat to Brocade than it is an advantage to the Director market and McData. The integration of SANs over IP backbones will be far more complex than current implementations and require more advanced software and processing capabilities, both McData strong points.
Fibre Channel over IP is an improvement that can be made to existing Fibre Channel networks but does little to alleviate any of the cost concerns and lack of widespread expertise associated with the technology. The second solution, iSCSI, offers a more drastic alternative. This solution proposes replacing the entire Fibre Channel infrastructure with an IP network, running over Ethernet in the LAN and IP over optics in the WAN. The argument against using IP for storage focuses primarily on three arguments. First, Fibre Channel currently supports higher speeds than most Ethernet LANs. Secondly, LAN traffic often involves variable delays, congestion and packet loss, three things that are not tolerable in storage networks. Finally, as I discussed in my report, IP packet processing has very high overhead since files must be broken down into small packets and then reassembled at the other end. This could potentially monopolize much of the systems resources and thus increase latency. ISCSI proponents believe that the combination of faster LAN speeds (ie. Gigabit and 10 Gigabit Ethernet) and improved Switched Ethernet networks will overcome much of the speed, congestion and packet loss issue. Furthermore, the iSCSI host bus adapter can be equipped with a TCP/IP protocol chip that can handle the packet processing without using significant system resources. An iSCSI network would be completely IP based and thus eliminate the need for the expensive Fibre Channel infrastructure. It would also have the ability to cross both LANs and WANs, hence eliminating the SAN island problem.
Both of these technologies are currently still in the process of being standardized so they have yet to prove themselves in the field. Fibre Channel over IP will likely be the first out the door but it is not as much of a complete solution as iSCSI which promises 100% IP. Both of these technologies are also still block based so they are unlikely to eliminate the File System advantage of NAS but they will definitely speed the convergence of the two technologies. A small startup firm called Nishan is also making waves in the market as it tries to get its Storage over IP technology incorporated into the official standards. Network Appliance is spearheading the DAFS (Direct Access File System) initiative based on VI (Virtual Interface) architecture, which hopes to bring many of the block level advantages to NAS. The primary advantage of DAFS is that it avoids the overhead of packet assembly/disassembly by bypassing the OS kernel and conducting direct memory-to-memory data transfers. This advancement should greatly increase the NAS value proposition and accelerate the technology's adoption in the high-end enterprise market, especially for database use. If this technology makes it to market, Network Appliance should be able to go head to head with virtually any SAN configuration, greatly expanding its addressable market. Since IP networks use existing Ethernet switches and routers, there would no longer be a need for Brocade’s equipment.
Conclusion
The danger posed by McData and Cisco along with threat to the entire industry created by IP based SANs and the Emergence of NAS appliances as an alternative architecture combines to create a rather bleak picture of Brocade’s future. If the world goes completely IP then it is unlikely that the company will be able to maintain its current level of dominance. However, despite the challenges ahead, Brocade is taking steps to prepare itself for this Brave New World. The company is addressing the problem by signing deals with companies like Nortel and ONI. Its most important partnership is with potential competitor, Cisco, which calls for the two companies to develop IP connectivity for the entire network (SAN, MAN and WAN) by incorporating a SAN interface into Cisco routers. Brocade currently has a great deal of muscle in this market as well as great management, research department and a war-chest of intellectual property, so it is highly unlikely that they will be blindsided by a company or technology any time soon. They should remain a major player but their position as a pre-eminent Gorilla candidate will likely not survive the transition. |
