Brocade showcased its 1860 Fabric Adaptor at Storage Networking World (SNW) in Orlando, Fla., this week, which gives customers the option to implement 16 Gbps Fibre Channel, 10 Gigabit Ethernet (10 GbE) or Fibre Channel over Ethernet (FCoE) connectivity. The company describes the adapter as “any I/O.” But Brocade product marketing manager James. D. Myers doesn’t see many companies  implementing FCoE so far.

“There isn’t a lot of adoption yet,” Myers said. “They are buying a lot of converged networks but they are not turning (FCoE) on yet. There are a few early adoptors. Most are hedging their bets.  I think it will take upwards of a decade for FCoE to be prevalent.”

Brocade hasn’t been a huge advocate for FCoE the way its rival Cisco Systems has been. But at least one SNW attendee confirms Myers’ thoughts.  Mitchel Weinberger, IT manager for the Seattle-based GeoEngineers,  said he researched FCoE and found the performance gain wasn’t significant enough to introduce a new technology into his infrastucture. The company uses an iSCSI SAN from Compellent that connects 10 GbE switches to virtual servers.

“We don’t see the benefit,” Weinberger said. “All the studies I’ve seen say the benefits are minimal. We really didn’t see enough advantage to put Fibre Channel over Ethernet. It’s another technology for us to learn, and we don’t have the staff.”

FCoE basically encapsulates Fibre Channel frames over Ethernet networks, and the benefits includes the reduction of I/O adapters, cables and switches in the data center. But the convergence of Fibre Channel and Ethernet means storage and network administrators must share management responsibilities, or one team must cede control to the other. That can be a big problem in organizations where the two groups don’t get along.

“It makes total sense,” said Howard Marks, chief scientist at DeepStorage.net. “Except for the politics.”

The storage networking vendor updated its product platform to 16 Gbps Fibre Channel this week, including a switch that supports FC and 10 Gigabit Ethernet (10GbE) ports to give it what QLogic calls “dual personalities.” QLogic also launched its 8300 Series Converged Network Adapter (CNA) that supports Ethernet, Fibre Channel over Ethernet (FCoE) and iSCSI, and the 2600 Series 16 Gbps FC HBA.

The Universal Access Point 5900 (UA5900) can be configured to run 16 Gbps Fibre Channel or 10 GbE traffic. Customers can start with 24 device ports and grow to 68 ports by adding licenses. Four of the ports can be used as 64 Gbps Fibre Channel trunking ports, and the switches can stack to 300 device ports. The UA5900 can be a Fibre Channel or Ethernet edge switch, and — with a Converged Networking license – it can serve as a top-of-rack FCoE switch to compete with Brocade’s 8000 and Cisco’s 5548UP devices.

QLogic also said it would bring out an intelligent storage router – called the iSR6200 – with support for Fibre Channel, FCoE and iSCSI. The router is designed for SAN-over-WAN connectivity.

The UA5900 and adapters are expected to ship through QLogic’s OEM and channel partners in early 2012, with the iSR6200 expected late next year.

QLogic was one of Cisco’s early allies in delivering FCoE gear years ago, and is on its third generation of converged networking devices. But FCoE has gained little adoption and Fibre Channel isn’t going away. QLogic execs say they expect Fibre Channel to remain strong while FCoE is a longer term item for many organizations. “We expect over the longer period, FCoE will gain momentum,” QLogic director of product marketing Craig Alesso said. “But Fibre Channel is still the workhorse for most enterprises.”

When FCoE does gain momentum, what role will hardware adapters play? Intel has launched software FCoE initiators that use host processing power and work with any network adapters. Intel’s plan is to eliminate the need for CNAs, but Alesso said QLogic’s adapters will have a big role in running FCoE. He maintains that CNAs are better suited for I/O processing and server CPUs should be used for applications.

“People can run FCoE initiators, but there’s a [performance] cost,” he said. “We free up servers to do what customers want to do with servers – run multiple virtual machines and multiple applications. The CPU should be used for running applications, not the I/O. We should run the I/O. Also, with [software] initiators, you lose management. You don’t have the common look and feel among management utilities.”

QLogic’s new products allow organizations to run FCoE, iSCIS and Ethernet traffic concurrently and use GigE and 10-GigE from one chip with its FlexLOM technology. John Spencer, QLogic’s manager of product marketing for its host solutions group, said he expects OEM and channel partners to sell the new products by the end of the year. Spencer didn’t name any partners, but QLogic’s press release quoted Dell, EMC, Hewlett-Packard and IBM.

QLogic also added features it calls ConvergeFlex, VMflex, FlexOffload, and SecureFlex. Ovbiously, the theme here is flexibility.

ConvergeFlex lets customers change protocols without taking down their servers.

VMFlex lets organizations carve physical ports into virtual ports and assign guaranteed bandwidth to each port. Customers can create four virtual ports per physical port and works with any 10-GigE Ethernet switch.

FlexOffload moves the workload from the processor to the adapter, freeing up bandwidth for virtual machines and applications.

SecureFlex encrypts data in flight over the network, and leaves encryption of data at rest to the storage array.

Wikibon analyst Stu Miniman said the best thing QLogic has going for it is expertise with all the protocols required for convergence while its competitors have had to add pieces of the stack.

“The biggest difference I see is QLogic has flexibility across Fibre Channel, Ethernet with FCoE and isCSI, and InfiniBand,” he said. “Today QLogic is the only vendor with the ability to run FCoE, iSCSI and TCP/IP concurrently. QLogic has more options than the competition.”

In a blog on QLogic’s new products, Miniman wrote the pieces for an all-Ethernet data center are emerging. However, he warned that convergence may be less of a money-saver than a process of merely keeping up with growing bandwidth requirements.

“Many people think that convergence will reduce spending but in fact it will only allow us to keep pace with the escalating server bandwidth requirements,” he wrote. “Wikibon estimates that in the next 72 months, servers will need close to 100X the bandwidth of today but traditional methods of delivering that capacity will fall far short of requirements. We believe that the market will need 6X what Moore’s Law alone will deliver. As a result, users should plan on buying more, not less equipment and they need to endeavor to find ways to reduce I/O consumption using techniques such as compression.”

InfiniBand providers claim their value proposition is the same as Fibre Channel over Ethernet (FCoE) — the ability to combine multiple types of network traffic over one wide pipe. But according to Brian Sparks, senior director of marketing communications at Mellanox and co-chair of the IBTA’s Marketing Working Group, RDMA over Converged Ethernet (RoCE, pronounced “Rocky”) can theoretically be run over DCE alongside Fibre Channel traffic. “With priority flow control in DCE, you can do both, but it’s up to the NIC provider,” Sparks said.

Sparks said Mellanox is planning to release products that support RoCE, and that other members of the IBTA manufacturers including Broadcom, Intel and Chelsio, may be working on 10-gigE NICs that support RoCE as well. The IBTA envisions RoCE finding use cases in financial services companies or businesses with a similar reliance on databases and low tolerance for latency.

IBTA is also hoping this spec will fare better than its last attempt to drive RDMA into the Ethernet world, a spec for RDMA over 1 Gigabit Ethernet iSCSI links called iSER, known in its Intel incarnation as iWARP. “It’s had a hard time getting market adoption, since the adapters are a little high on the power side,” Sparks said. “They require a slightly more expensive NIC and maybe only get a microsecond or two latency benefit.”

RoCE will be different, Sparks said, because latency is much improved over 10-Gig networks, and DCE will do away with the need to offload TCP/IP traffic, one of the hindrances to better performance with iWARP.

The Virtual Fabric Adapter is a virtual Network Interface Card (NIC) that goes into each blade server in the BladeCenter chassis. Rather than the typical four to six separate IP ports and two Fibre Channel ports, the VFA will run Gigabit Ethernet and Fibre Channel over Ethernet traffic streams over four 10 Gigabit Ethernet ports.

Emulex has also added support for Blade Networks’ 10 GbE switch, which connects the BladeCenter to the wider corporate network.

IBM previously announced it will also resell converged networking products from Cisco, Brocade and QLogic. IBM officials also said at the time of that announcement there were plans to add products from Emulex as well, but specifics were not made public until this week.

NetApp is the only storage vendor to announce native Fibre Channel over Ethernet (FCoE) interfaces on its disk arrays so far. EMC Corp. and IBM partner LSI Corp. have said they don’t expect significant traction for FCoE in storage arrays until next year. Analysts say they don’t expect FCoE to reach a ‘tipping point’ over Fibre Channel until at least 2011.

Jeff Boles, director of validation services and senior analyst for the Taneja Group –

Right now, we have a scattering of fabrics and technologies, and while the promises of FCoE are interesting (if not compelling) for the day to day practitioner, transitioning to this new fabric is a bit more complex than filling your shopping cart on Amazon.com.

What I fully expect to happen is a multi-year integration of converged Ethernet as a broad fabric that joins together the multiple fabric domains in the enterprise data center. Those separate domains – FC, InfiniBand, and even traditional Ethernet – may rapidly become converged in a 10gb core, but will likely keep growing at a steady pace, or at least being maintained with regular equipment replacements. Once a converged core is in place (over years), we’ll likely see new equipment deployments taking place on the converged fabric when it is justified (high I/O demands, cable simplification in large infrastructures).

But a full tilt shift to FCoE as the new fabric, is likely out beyond the 3 year mark for aggressive businesses, and well beyond the 5 year mark for less aggressive businesses. The problem plain and simple, is that many, many businesses are well served by their current fabrics and skillsets, and the transition to converged ethernet, and FCoE, will only get near term adoption when it is fully justified. Many times, existing fabrics and skillsets will outweigh the battle over port prices and power utilization. While CEE/FCoE will change the computing landscape, my expectation is that this will happen in the long term. 

Andrew Reichman, senior analyst for Forrester Research –

I’m seeing vendors like Brocade, Cisco, QLogic and NetApp move towards greater support for FCoE. The benefits often include reduction of complexity in cabling, and a longer term desire for simplification of SAN and LAN networking through network convergence. That said, it is likely to take a long time to see the benefits, and require a fairly significant investment in new equipment and re-architecting. I do believe that storage traffic will be on Ethernet at some point, the question is how soon- The FCoE standard has been slow to emerge, which has delayed adoption, but early adopters seem to be getting started now. 2011 seems a bit ambitious for broad adoption beyond FC, but I think it might not be too far off. You do have to remember that storage buyers are extremely conservative and like to see very mature products and architectures before making a big change, but once the momentum gets going, it’s likely to grow rapidly.

Mark Kelleher, Managing Director, Equity Research, Brigantine Advisors –

Dell’Oro isn’t really going out on a limb with its prediction that FCoE will supplant Fibre Channel by 2011 – that’s a common assumption in the storage industry. One converged fabric for all enterprise communications makes a lot of sense. The fibre channel switch and HBA people are moving in that direction, the Ethernet providers are moving in that direction, there’s really no reason it would not. The key difference between FC and Ethernet is that Ethernet can lose packets and take its time to recover, while FC guarantees delivery, and does not drop packets. To port the upper layers of the FC stack onto Ethernet, the Ethernet protocol itself has to be augmented to allow ‘lossless’ transmission of data under certain circumstances. That is all incorporated in FCoE, and the technology is just now reaching the market. Deployment starting now thru next year, widespread adoption by 2011.

Keep an eye on the core FC vendors: Brocade, Emulex, and QLogic. Brocade sells switches (although moving into the host-bus adapter market), while Emulex and QLogic are knows for selling the input/output offload engines that connect servers to FC (host-bus adapters, or HBAs). To connect to Ethernet, servers use “Network Interface Cards”, or NICs. With the new FCoE protocol, those two functions are combined into a “converged network adapter”, or CNA. Sell through of CNAs will tell us how the adoption of FCoE is progressing.

Reinoud Reynders, IT manager at University Hospitals Leuven in Belgium–

I believe very strongly in FCoE. Cisco is pushing this very hard and indeed, they have a strong story. Just one plug for all your I/O (network and SAN) on 10 Gb, 1 switch that separates client access (IP network) [from the] storage network: it’s a great plan.

I will replace my FC-SAN switches [around] Q2 2011. Personally, I believe 2011 is a little bit to early for the [broader industry] cross over, but maybe 2012.

Feel free to add your own perspective in our comments section below!

After Emulex repeatedly spurned its acquisition offers, Broadcom today threw in the towel and said it will walk away. However, the Ethernet-chip maker may not be giving up on buying a storage company altogether.

“Although we were unable to negotiate an expeditious and friendly transaction at a price that makes sense to us given the expectations set by the Emulex board, there are other value-creating alternatives that we will now turn our attention to as we position Broadcom to capitalize on the emerging opportunities in the converged enterprise networking markets,” Broadcom CEO Scott McGregor said in a news release.

Broadcom’s decision to walk came after Emulex’s board today rejected its latest offer to buy the HBA company, claiming the bid of $11 per share or about $912 million is still too low.

“We unanimously believe Emulex will deliver significantly more value than Broadcom’s revised offer through the company’s rapidly developing converged networking business and solid execution in our host server and embedded storage markets,” Emulex chairman Paul Folino said in a statement.

Folino did add, “we would of course give full consideration to a bona fide offer from any party that reflects the full value of the company.”

Broadcom had been chasing Emulex since December, and made its first formal offer of $9.25 per share or $764 million in April. Emulex management has accused Broadcom of trying to take advantage of its depressed stock price during bad economic times. Emulex shares closed at $9.70 Wednesday, but opened at $8.44 today and traded at $8.90 by mid-afternoon after the deal fell through.

What are Broadcom’s alternatives? Emulex rival QLogic would be even more expensive to buy, so that’s probably not an option. Broadcom can try to build its own Fibre Channel over Ethernet (FCoE) products, or it could pursue LSI’s dormant FC HBA technology.

Emulex also said today it expects to report revenues of approximately $78 million to $79 million for last quarter, in the high end of its forecast of $73 million to $80 million. CEO Jim McCluney added that Emulex recently scored two OEM design wins for its LightPulse HBAs and two design wins for its 10 Gbps Ethernet OneConnect converged network adapters – all with tier 1 vendors. Emulex did not name its partners on those deals.

Since Broadcom went public with its move on Emulex last week, Emulex executives have been positioning their company as one with enough Ethernet capability to successfully compete with Broadcom and others on that front.

The new spin started on the Emulex earnings call Monday night. “The cornerstone of our strategy is the converged data center, based on 10-gig Ethernet technology,” CEO Jim McCluney said. “Our converged data center network is one that unifies IP and storage networking over a single wire.”

McCluney said the strategy revolves around its OneConnect Universisal Converged Network Adpaters and OneCommand management platform. Both rely on 10-Gigabit Ethernet for iSCSI, Fibre Channel over Ethernet (FCoE) and enhanced Ethernet.

With ASICs it gained through a joint development partnership with startup ServerEngines, Emulex claims 16 10-Gigabit Etherenet-based cards, five 10-GigE NIC design wins, three 10-GigE iSCSI adapter wins, and four Fibre Channel over Ethernet (FCoE) design wins with OEM partners.

“While we’re not in a position to announce the details of these wins, we believe some of them have come at the expense of our leading competitors, including Broadcom,” McCluney said.

Financial analysts on the call got the point. “On numerous occasions during the earnings call, Emulex alluded to 10GbE, iSCSI, and FCoE-based, Tier 1 OEM design wins, suggesting the company may not require additional Ethernet-based expertise to compete well in the ongoing unified fabric adoption cycle,” analyst Amit Daryanani of RBC Capital Markets wrote in a note to clients.

So when I spoke to Emulex chief marketing officer Steve Daheb this week about the company’s strategy, it was no surprise he declared Emulex an Ethernet company.

“People are saying,’ You’re kidding me, you guys have done the Fibre Channel thing, now you tell me you’re winning 10-gig NIC deals, Ethernet-based deals?’” he said. “And we are.”

Daheb added that Emulex isn’t abandoning Fibre Channel. It will add encryption and security features to its HBAs, and will support the FC roadmap beyond 8-Gbps.

“We continue to invest in Fibre Channel,” Daheb said. “We have 8-gig HBAs today, and we’ll be a player in 16-gig Fibre Channel.”

But Emulex sees there is more to the world than FC. It could even conceivably follow its HBA rival QLogic into InfiniBand.

“It’s something we’re watching carefully,” Daheb said. “We see InfiniBand for inter-switch clustering catching on. But we’re betting on Ethernet here. With the low latency [of enhanced Ethernet], you get a lot of those same benefits as InfiniBand.”