Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Gorilla and King Portfolio Candidates -- Ignore unavailable to you. Want to Upgrade?

To: Bruce Brown who wrote (19543)	3/8/2000 10:48:00 PM
From: Greg Hull	Read Replies (3) \| Respond to of 54805

RE: Brocade Bruce, Speaking of Brocade, I think they may have disclosed today how they intend to make a Gorilla Game out of what appeared to me to be a Royalty Game. They announced this morning plans to have their routing protocol blessed by ANSI. biz.yahoo.com First, some background for those who don't follow the Fibre Channel switch market closely. A typical FC switch has 16 ports. Usually computers or storage systems are plugged into the ports on a switch. These ports are then used as FL_Ports (Fabric Login). It is possible to plug two switches together, however, as a way to have more devices on a single SAN (Storage Area Network). In this case, one port on each switch is used as an E_Port (Extension Port) instead of an FL_Port. Now only 15 computers or storage devices can plug into each switch, but the total SAN can have 30 devices rather than just 16 devices. It is possible to string several switches together this way to construct a SAN of any size required. This daisy-chain topology is possible, but not preferred. If switches are stitched together serially, a device plugged into Switch #1 has to hop through 9 switches to connect to a device on Switch #10. While each port can communicate at 1 Gbps (or faster in the near future), the latency increases for each hop made. [For those who remember the analogy I gave a couple weeks ago, these hops are equivalent to mailing a newspaper to City A, then mailing it to City B, then ... to City J. When the recipient in City J finally received the newspaper, the information would be many, many days (weeks?) old.] A better way to assemble a large SAN would be to add an eleventh switch to the fabric. Instead of stitching each switch together end-to-end, each of the 10 switches would plug into Switch #11. Now the worst case number of hops (latency) in reduced from 9 to 2. (hop from Switch #1 to Switch #11, and then hop from Switch #11 to any of the other switches 2-10). This is referred to as a Cascade Topology. There are other topologies for large SANs that add multiple switches for redundant data paths so that a failure can not take down the whole SAN. This is analogous to having more than one fire escape for a building. If there were only one escape and a fire broke out near it, people might be trapped inside. Redundant escapes reduces the chance of a catastrophic failure. Today multi-switch fabrics are possible if the switches are made by the same vendor. The Fibre Channel standards address many issues, but not E_Ports. There have been several efforts to address interoperability of switches (and hubs and other devices) such as the FibreAlliance headed by EMC and the Jiro project headed by Sun Microsystems. Brocade has not joined either of these efforts. Today Brocade announced that it will contribute its routing protocol, called Fabric Shortest Path First (FSPF) to a standards committee "that will facilitate baseline interoperability among Fibre Channel switches from different vendors". Note the word baseline. Brocade intends to release "certain elements" of its routing protocol. I'm interested in hearing other interpretations, but mine is that Brocade intends to have an open, proprietary architecture (FSPF) that builds upon the open, non-proprietary Fibre Channel standard. Brocade has not joined the FibreAlliance and Jiro efforts because they would not own the inter-switch communication protocol. One would fully expect Brocade's implementation of FSPF to have additional features and capabilities. Further, Brocade would offer these additional features to the standards committees from time to time as necessary to maintain control over the protocol. BTW, a large SAN does not necessarily require a large number of switches. By using a switch with 64 or 128 or 256 ports, the need to stitch or cascade switches is eliminated. As buck has pointed out, these high port-count switches are usually called directors, but they function the same as low port-count switches and have high bandwidth backplanes and redundant data paths. While the large SAN described above required 11 16-port switches for the cascade topology, the 150 devices on it could be handled by only one 256-port director. Further, the SAN with the director would have lower latency. From the recent IDC market study "the director switch is the fastest growing segment in terms of revenue. Director class switches are increasing at a rate faster than all other segments combined -- a compound annual growth rate (CAGR) of 129 percent between 1999 and 2003. In addition, the director class switch will take over fabric switches in 2001, and by 2003, they will account for 52 percent of the market's revenue." Merlin suggested in our discussion last month that there could be two Gorilla Games present: one for low-count (fabric) switches and one for high-count (director) switches. Brocade's strength today is in the fabric switches, and they play little or no role in the SAN director market. If the SAN fabric switch market enters a tornado (and I think it will) and Brocade succeeds in having FSPF displace the FibreAlliance and Jiro attempts at switch interoperability, it sure looks like Gorillahood to me. Greg