The media likes flashy headlines that talk about a new technology destroying an old one. When it comes to Fibre Channel over Ethernet (FCoE), the number one reason for creating the technology was not for the storage vendors to expand the money that can be squeezed out of customers or the network vendors killing FC. The primary reason to create FCoE was for server vendors that required a single network to create next-generation, dense, blade servers that are optimized for virtualization. Power and density requirements could no longer support multiple protocols on separate adapters and switches inside the chassis. FCoE allowed Cisco to bring its UCS platform to market and is also used in the latest generation blade servers from IBM and HP. Bottom line, this first mission of FCoE has been accomplished. Customers don’t need to “want” or “need” FCoE for this first piece, LAN and SAN management models remain untouched.
As for moving to an all-Ethernet environment, everyone would love to wave a magic wand to move to a single network, but that’s not reality. While NetApp, Compellent, and EMC provide native FCoE storage arrays, storage customers are notoriously conservative, and a full transformation to a converged environment (and the organizational and operational impacts) will take time. Plenty of solutions are available for FCoE; the products that I believe have the best chance of helping customers through the next phase of convergence are the ones that not only work with existing environments but also move customers towards the desired end-state of a single network while providing complete investment protection and flexibility. This means that they can be deployed to work with existing FC environments and repurposed to be fully Ethernet in the future. Today, Cisco, HP, and Juniper offer products that can switch between Fibre Channel and Ethernet without changing hardware.
HP was first to market with ports that could support FC or Ethernet without changing hardware. HP calls this product “flex ports” and ships it in blade servers (using QLogic technology to create the HP FlexFabric Virtual Connect), rack-top switch (A5820) and is expected to move this into core switches later this year. A proof point of HP’s push into the FCoE market is Emulex claimed market leadership in Q4’10 10GbE LOM ports according to Dell’Oro; a solution that HP OEMs. As a quick aside, in the overall 10GbE adapter space (including FCoE), according to Crehan Research, for CY10, Intel shipped 221k ports, QLogic 163k, and Emulex 118k. It’s a heated battle for LOM design wins and customer loyalty in the adapter space.
Cisco has “unified ports” – first shipped on the Nexus 5548 (1U 48-port switch shipping now) in the expansion ports – a 16-port module that supports 1/2/4/8 FC or 1GbE/10GbE/10GbE with FCoE. While unified ports do not currently support dynamic switching of FC to Ethernet, the ports can be switched without changing any hardware; it simply requires an offline change to the ports. Cisco also expects to ship the Nexus 5596 (2U 96 port switch) in Q1’11, which will also support the unified port functionality.
According to Cisco, approximately one-third of all Nexus 5000 switches have FCoE licenses enabled. Those who think that Cisco support for iSCSI shows a lack of faith in FCoE are wrong. At the end of the day Cisco wants everything to run on Ethernet, and both iSCSI and FCoE help storage customers get there.
It’s a small piece of the QFX3500, but it does support functionality similar to flex/unified ports. The QFX3500 is a 1U 48 port top-of-rack switch – all 48 ports support 1Gb or 10Gb Ethernet and an additional 4 ports are 40Gb Ethernet for attaching to interconnect chassis (as part of Juniper’s Stratus Project that delivers a fabric of switches as a single tier). Of the 48 ports, 12 of the ports can also be 8Gb FC, and similar to the HP and Cisco products discussed above these ports can be switched as needed to be either FC or Ethernet. Juniper developed this solution in-house, and it is its first product with FC. According to Juniper, it has done extensive internal tests with IBM, NetApp, Brocade, and Cisco products. In discussions over the last three years, customers looking for a second source for FCoE technology have looked for Juniper to be that option.
What about Brocade?
Brocade is the leader in FC and has a full line of FCoE products, but none of them currently support the ability to switch ports between FC and Ethernet. More troubling for Brocade in the space is that its executives have been quoted as being somewhat skeptical of converged networking. For instance, Brocade's CMO recently said that FCoE may not happen. The issue for Brocade is that it is not seeing the same volumes of FCoE sales as HP and Cisco. Brocade does not have any embedded design wins, and the FCoE offerings (like the Brocade 8000) are separate from the Ethernet products (from the Foundry acquisition). A couple of good quarters of FC sales is not a reason to ignore the opportunity of FCoE. Brocade’s quandary is that providing a smooth path from FC to Ethernet where the FC can disappear is scary for the market leader, but if it won’t offer the solution, the competition already does. Emulex and QLogic have done a good job on maintaining FC adapter sales while competing with Intel and Broadcom for Ethernet business.
Action Item: CIOs are not looking for new protocols or even “convergence”, they are looking for solutions that simplify operations, save money, and allow for flexibility in an uncertain future. Embedded solutions of FCoE and switches with flex/unified ports that can span both FC and Ethernet solutions deliver on these requirements. Storage networking vendors fall into two camps, those delivering FCoE solutions to customers and those saying that there is no market for it.