McData Wants Brocade to Disable Hardware
By Tim Stammers and Rik Turner
The two giants of SAN switching entered legal combat last week when McData Corp launched a lawsuit apparently aimed at the jugular vein of Brocade Communications Systems, in an effort to blunt a coming assault on its territory.
Alleging a patent infringement, McData says it is determined to force Brocade into disabling key frame filtering functionality within its products, something that could be a major engineering task for its rival company. The suit comes at a critical time for Brocade, and was launched the day after it finally announced that it is poised to ship a key product - its Silkworm 12000 with which it hopes to challenge McData at the high end of the market. That device is already late and Brocade can ill afford to see it slip any further.
The happy synchronicity of the suit, and the fact that a McData spokeswoman said she is "not aware of any plans" to launch suits against other companies also involved in frame filtering such as Rhapsody, Confluence, and Maranti, may suggest that the suit is only a feint or delaying tactic. Brocade suggested as much last week, and said it will fight the action "vigorously." It has however already settled one patent lawsuit from McData out-of-court, with a cross-licensing deal signed in 1998.
McData declined to make any statements about the timing of its action."Obviously law suits take a while to file. I can't comment on whether it [the timing] was a coincidence," the spokeswoman for McData said. McData - a former subsidiary of the famously aggressive EMC Corp - is however playing it to the hilt and has declared that it is after Brocade's blood, and is not prepared to accept damages, or negotiate a licensing deal. "The outcome we're after is disablement of Brocade's products," a spokeswoman for McData said.
That could be very damaging to Brocade. Not only would that involve the SN12000, but also Brocade's SN 38000 mid-range switch, the only shipping hardware from the company to implement the latest 2Gbit version of Fibre Channel. Last week Brocade predicted that these two products will account for 60% of its revenues in its current fiscal quarter.
Frame filtering is an emerging and valuable facility, but the real significance of the lawsuit is in the difficulty of removing it from Brocade's products. Brocade is likely to have implemented it in silicon, as McData has done in its initial implementation of the technology. Brocade was last week unable to confirm whether it has chosen the same route, but admitted that the common factor between the 3800 and the 12000 is the ASIC switching silicon.
"It would delay the SN 12000 by a minimum of six months, unless there's a way of turning off the frame filtering programatically. Otherwise it could be a huge deal," said Arun Taneja, analyst at the Enterprise Storage Group.
Brocade and McData are the two juggernauts of Fibre Channel SAN switching. They have yet to really cross swords, but are set to do so. Each holds around 90% share of their respective markets. For Brocade, that is the market for mid-range switches, and for McData it is the market for high-end large and sophisticated switching devices, known as directors. However Brocade's market is being marginalized. By its own estimates corporate SANs are growing at an average increase of 100 ports per quarter, and as a result its mid-range products are being renamed by some as simply "edge switches."
McData already offers these devices, albeit rather tamely, unveiling 16- and 32-port offerings in the fourth quarter of 2000 but, to date, has cornered only 8% of the midrange market by its own admission. If Brocade is to protect its stronghold there, it too needs to offer an end-to-end solution. A delay to the 12000 will give McData more time to whittle away at Brocade's market share. The device was originally promised to ship last year, and rumors that it was suffering development problems attracted a lot of attention from Wall Street last year. Last week, Brocade dismissed the rumors and said it will ship in this quarter.
As to how serious a challenge Brocade can mount to McData in the director class, Brocade itself is careful not to refer to the 12000 as a director, not least because its architecture is different, essntially comprising of a multiple of smaller boxes joined together (in fact, it is two of the 64-port switch which, in turn, is really four of the 16-port ones on a blade). It clearly will not debut with all the RAS functionality characteristic of true directors but, as the pundit put it, "some customers for Brocade's low-end switches want to stick with the same supplier as they grow, even though true directors have much greater functionality."
Beyond that, the issue is one of OEMs. EMC is still McData's primary customer in that area, say industry sources, with several of the other big players reselling Inrange, so in some sense, if Brocade can leverage its relationships with those other server vendors for its 16-port switches into its higher-end offerings, they might start reselling the 12000 alongside, or even instead of, the Inrange products, giving Brocade greater collective firepower vis-a-vis McData, albeit with a less performant switch.
McData denied the claim of one ComputerWire source that prior to its spin-off from EMC in 2000, it had completed joint research with Brocade. This the source said ended with a divorce settlement that divided up the resulting IP, and barred either side from suing for infringement of patents. McData said this version of events may have resulted from confusion over the April 1998 cross-licensing deal signed with Brocade.
The source said that even it such a joint development deal were signed it would be very unlikely that Brocade would be stupid enough to welch on it. Indeed, he went on, "this is almost a publicity stunt", opining that "McData still have that EMC mentality of putting winning first. They'll do anything to win business."
computerwire.com
Patent Number 6,233,236 has 13 patent references. The patent background provides a description of the problems that the patent is trying to solve and the limitations of the prior art, i.e., the 13 patent references. The rest of the patent establishes the novelty and the non-obviousness of the patent in several preferred embodiments.
Put simply, if the Fibre Channel open standard is A+B and this patent is A+B+C, Brocade has to come come up with something like an A+B+D patent to avoid infringement, but McDATA has 25 utility patents and 15 pending patent applications after $146M spent on R&D in the last 5 years so most likely McDATA has patents covering any combination of A+B+C+D or A+B+D+E etc. That leaves little room for Brocade to maneuver and explains why after $200M spent on R&D in the last 5 years, it has only produced 1 utility patent and a handful of patent applications that keep on getting shot down or delayed at the USPTO by one McDATA patent or patent application after the other.
Put another way, this patent provides the basic statistical eyes for the management consoles at the device level, at the fabric management level and at the enterprise level so if Brocade is forced to disable this ASIC-level feature, its SANs will continue to be blind. Brocade SANs are already notorious for being unnecessarily complex and highly prone to latency with SAN expansion. That advantage is going to snowball against Brocade when McDATA starts to introduce more advanced fabric services on top of that statistical foundation that facilitate the same type of storage-server, storage-storage and server-server traffic in the mainframe ESCON/FICON world.
From the '236 patent......
......A switch is a multi-port device where each port manages a simple point-to-point connection between itself and its attached system. Each port can be attached to a server, peripheral, I/O subsystem, bridge, hub, router, or even another switch. A switch receives a connection request from one port and automatically establishes a connection to another port. Multiple calls or data transfers happen concurrently through the multi-port fibre channel switch.
A key advantage of switched technology is that it is "non-blocking" in that once a connection is established through the switch, the bandwidth provided by that connection is not shared. Hence, the physical connection resources such as copper wiring, fiber optic cabling and radio frequency bandwidth can be more efficiently managed by allowing multiple users to access the physical connection resources as needed.
Switched communication systems work most efficiently when the switch resources are allocated efficiently. In other words, in a multi-port switch, each of the ports desirably processes an equitable amount of data traffic. When one connection through the switch absorbs an unbalanced amount of traffic, traffic delays are caused even while the switch possesses unused resources (i.e., unused ports and connection resources). The end result of an unbalanced switch is unnecessary average latency in processing traffic caused by bottlenecked connections within the switch.
Users desire knowledge about the data traffic that flows through a switch in order to efficiently manage that data traffic and avoid bottlenecks. A switch is most efficiently used if the traffic load through that switch is reasonably balanced. When a particular switch port or ports are used heavily by a single host or destination those ports are not available to handle other traffic. Similarly, it is desirable to have knowledge about the type of data traffic flowing between a pair of ports. In a fibre channel system, for example, the data traffic may comprise any of a number of FC-4 specified frame types. Knowledge of the frame type can be used to configure and reconfigure the communication network for more efficient utilization of resources.
Prior approaches provide a means for counting received or transmitted frames on a single port basis. This allowed a user to monitor traffic at a single switch node, however, does not provide an mechanism to reliably correlate the traffic between a source port and a destination port within the switch. In other words, the prior approaches enable one to monitor the traffic from a node port to a switch port, but not from one switch port to another switch port. The prior approach enables one to determine if a particular path or port is experiencing an unusually high load, but does not provide sufficient information to determine what host(s) is/are driving the port. If multiple hosts are driving the port, prior approaches are unable to identify the proportionate share of the load of the port for which each of the driving hosts is responsible. Moreover, because prior approaches are not protocol specific, it is difficult or impossible to tell what type of traffic is consuming the bandwidth of a particular host port.
A need exists for a method and device for measuring. protocol traffic that provides information about traffic within a switch from a source port to a destination port. A need also exists for an analysis tool that measures traffic on a protocol specific basis........
patft.uspto.gov
Contributory Infringement (sale of a component):
In summary, to be a contributory infringer under 35 U.S.C. 271(c), a party must sell a component used in a patented combination, where:
the component constitutes a "material part of the invention";
the component has no other "substantial, noninfringing" uses;
where the party knows the component to be "especially made or especially adapted for use in an infringement of such patent."
The contributory infringement statute is designed to prevent those who knowingly provide a specially adapted component to be used in an infringing system from profiting from the sale of that component. The contributory infringement statute has very important exceptions that will not allow the broadening of the patent right to cover the sale of unpatented staple articles of commerce, even if they are used in an infringing process. It is also important to note that there can be no contributory infringement in the absence of direct infringement......
google.com |
