Article originally posted in May 2011
Convergence or FC vs Ethernet
Next week at Interop in Las Vegas, there will be a barrage of press and demos including potentially nine new 10Gb Ethernet switches (according to Nick Lippis) and likely a round of 40Gb Ethernet announcements. Today, Emulex made the first product announcement for 16Gb Fibre Channel (FC). Some will write headlines about the battle of FC vs. Ethernet and how 40GbE and 100GbE obviate the need for new 16Gb FC or the creation of 32Gb FC in the future. For the vendors that play in both FC and Ethernet markets, it is a tough to manage the balance of products and messaging. The switch vendors do not do a good job at this balance, Brocade does not move aggressively enough towards convergence, and while Cisco is a solid #2 in FC switching, marketing is always focused on Ethernet. Adapter and chip vendors QLogic and Emulex lead the way towards flexible converged solutions that allow FC and Ethernet to be complimentary.
Adoption of Speed Increases in FC and Ethernet
Before we look at 16Gb FC, it helps to understand how the adoption of FC and Ethernet take place. FC upgrades are in 2x increments, now on its fifth generation in the last dozen years (1, 2, 4, 8 and now 16Gb FC). From the time that a new speed is ratified at the standard committee, it usually takes about three years for a sizable portion of new sales to be on the new generation. The speed of adoption on switches is a little faster, since the cost and performance of new speeds make sense for inter-switch links, and the switch vendors can push customers to the new generation through pricing and end-of-life of previous generations. On the adapter side, OEMs typically tie the release of a new generation adapter to servers that upgrade the server bus (PCI, PCI-x, PCI Express, PCI Express Gen 2, etc). It is fairly common for customers to skip a generation of FC speeds since the technology adoption curve is faster than most customer equipment refresh cycles. Speed increases of FC are typically similar in price to previous generations.
Ethernet has historically increased speed in 10x increments with adoption taking 7-10 years for significant volumes after the standards are complete. Similar to FC, new Ethernet speeds start in backbone deployments, work through the switching layers and then into servers. Equipment refresh cycles have often lined up with new generations of technology. The price, power consumption and even cabling requirements can slow the adoption of new generations of Ethernet.
Enter 16Gb FC
The standard for 16Gb Fibre Channel was ratified in September 2010 (see the FCIA press release). FCIA claims that 16Gb FC will deliver twice the bandwidth and consume 25% power.
The most important early adopters of new interconnect technology are the OEMs. For embedded adapters, it typically takes 12-18 months to move from concept to production. Once OEMs start shipping a product, customers typically need to qualify the new device, verify all necessary interoperability and even the simple process of changing which SKU purchasing should order can take quarters or years for larger companies.
From a technology standpoint, most server manufacturers are going release 16Gb FC with the PCI Gen 3 bus, probably later this year. This enables full-speed 10Gb Ethernet and 16 Gb FC. While server virtualization and cloud computing are increasing the demand for high speed interconnects, only a minority of customers are not currently satisfied with 8Gb FC or 10Gb Ethernet options. Therefore, performance will not be the decision point for most customers choosing between FC and Ethernet. Users that are interested in Ethernet will find the “40% faster than 10Gb Ethernet” claim of 16Gb FC insufficient to move people back to FC. Similarly, the promise of 40 and 100Gb Ethernet is unlikely to push customers that want to stay with FC away from 8Gb today, 16Gb soon or 32Gb in the future. The options of FC and Ethernet are converging thanks to FCoE and flex/unified ports.
|Rating||Actual Data Rate||Real Throughput|
Thanks to Cisco’s J Metz providing this data.
Emulex does more than 16Gb FC
Emulex is the first to announce a 16Gb FC product – the XE201 controller, which will be called the LP16000 when put on a server adapter. What makes the product more interesting is that it can support 8Gb FC, 16Gb FC, 10Gb Ethernet or 40Gb Ethernet (including combinations such as 1x16 FC with 2x10GbE or 2x8Gb FC with 2x10GbE), making it a combination of a traditional FC HBA and the newer Ethernet-based CNAs. From a power standpoint, the new adapters have a power rating of 11-15W; a bit higher than single-port FC or 10GbE, but quite low for a single-port 40GbE or quad port 8Gb FC. While most customers typically stick with a single configuration (and protocol), the flexible options should make the XE201 chip attractive for embedding in OEM solutions. The LP16000 is not available for purchase yet; according to Emulex the adapter is expected to ship towards the end of 2011. The release will probably be tied to an OEM release of product, and Emulex has not announced any design wins yet. EMC participated in Emulex’s webcast, but storage is typically the last to market, and unlikely to ship native 16Gb FC until late 2012 or 2013.
Competition for 16Gb FC
The FC adapter market is a duopoly that is more focused on the transition to Ethernet than on battles in FC. As Emulex is not shipping product until later this year, and adoption takes a long time, the real battle between Emulex and QLogic in the FC market is about OEM design wins more than product announcements. QLogic is the market leader in FC, and while Emulex’s technology looks good, changing vendors involves a significant cost for OEMs. Some vendors are starting to provide Modular LOM (see discussion of QLogic FlexLOM which Dell started shipping last year) that allows OEMs to offer multiple vendor options on a single platform. End-users may not know the difference between adapters and will often simply go for the cheapest widget.
What about the switches? At SNW, Brocade stated that it will ship 16Gb FC adapters and switches in the first half of 2011. UPDATE: Brocade did announce 16Gb FC core directors, edge switches and Fabric Adapters on May 3rd (see PR). Cisco was almost a year behind Brocade shipping 8Gb FC, and while I do expect that Cisco will make a 16Gb FC MDS switch, it will likely try to freeze the market with Ethernet until that product is available. Since most customers will not put 16Gb FC or 40GbE into production environments until 2013 or later, any vendor sniping can be safely ignored.
Action Item: Most users can take a slow and deliberate approach to the adoption of new generations of speeds. While changes are a time to evaluate potential changes to architectures, thanks to technologies like the Emulex LP16000, customers can support both FC and Ethernet and consider the migration on internal schedules rather than on the pace that the vendor community may want to push or pull. For equipment refresh cycles that start in 2012 or later, consider looking for adapters that can support the latest of both FC and Ethernet.
Footnotes: Stu Miniman is a Principal Research Contributor for Wikibon. He focuses on networking, virtualization, Infrastructure 2.0 and cloud technologies. Stu can be reached via email (firstname.lastname@example.org) or twitter (@stu).