Three weeks ago I wrote about who was filing patents around the emerging FCoE standard and noted the curious absence of Brocade.

So much for the lack of filings by Brocade- the USPTO has just published an application by Brocade Communications (application 20080159260).

As the FCoE standard moves closer to completion, companies are beginning to make product announcements. The most interesting development so far is not the announcements by the expected players such as Qlogic and Emulex, but also the stated lack of support by Microsoft.  Both Qlogic and Emulex are trying to maintain their duopoly in the FC HBA market by introducing combo FC and FCoE Ethernet CNAs- converged network adapters.  Both vendors products are expected to integrate seamlessly into existing SAN infrastructures.

But FCoE does not require a dedicated HBA like Fibre Channel.  FCoE uses Ethernet and can be enabled with a software driver.  This raises the question of performance, but with the abundance of processing available from multi-core processors, performance should not be an issue for FCoE initiators.  Dedicated FCoE targets (when available) may use hardware HBAs.

There is a good chance that the FCoE HBA market may go the way of the iSCSI CNIC market- software from Microsoft and Intel (via the Linux openFCoE project) for most initiators using standard Ethernet NICs with hardware FCoE NICs used only for targets.  This will make it tough on the HBA vendors- The target HBA volume is a fraction of the initiator HBA market.

Several companies have filed patent applications with the US Patent Office for patent protection on technologies relating to the emerging Fibre Channel over Ethernet standard.  The applications have been filed by Cisco (application 20060251067, application 20060098681), Emulex (application 20080056300) and 54th Street Systems (application 20080028096).  All the patent applications were filed prior to the FCoE project start date by the FCoE standard committee.

Also interesting is the lack of filings by Qlogic and Brocade, both major players in the Fibre Channel SAN market.  Of course, patents with FCoE directly referenced are not the only areas where FCoE products can infringe on other patents.

Disclosure: I am CTO and one of the founders of 54th Street Systems.

As I mentioned in my recent post FCoE – Yet Another Storage Protocol?, FCoE needs to provide three things to be successful- It must provide a feature advantage, it must provide a cost advantage and it must integrate easily with existing storage infrastructure. This post focuses on the second item- The Cost Advantage.

FCoE is a technology designed for servers. It allows a server to use Ethernet for both networking (LAN) and storage (SAN) access. The first cost advantage of FCoE is using Ethernet which uses the existing infrastructure. Unlike iSCSI, the other network based SAN access protocol, FCoE does not require a TCP protocol stack to operate. The lack of a TCP stack reduces complexity and latency. The second cost advantage of FCoE is using a simpler host interface that does not need a TCP stack to access storage. The third cost advantage of FCoE is leveraging existing networking management knowledge and training.

The combination of these three advantages; Ethernet, reduced complexity and leveraging existing management kowledge and training enables FCoE to become a viable competitor for host SAN access.

Until now, the industry has been looking to the blade server vendors to provide management tools for blade servers. On the surface, this looks like a reasonable expectation. But I think this shows a gap in thinking.

One of the goals of managing blade servers is managing each chassis or rack as a system, instead of managing each blade individually like a rack of 1U servers. This type of management is a complex task. Managing blade servers as a system requires managing multiple operating system instances and their associated resources. These are also the same requirements that virtual machine companies have for the management of virtual machines.

The companies that must accomplish this are not the blade server vendors; the blade server vendors will survive if they provide system management tools or not. The companies that must accomplish this are the virtual machine vendors. The VM vendors, by the definition of their business, manage multiple instances of operating systems with some resources being shared between the instances. If the VM vendors do not provide systems management for VMs, their business will fail.

The companies who’s business are dependent upon effectively managing multiple instances of operating systems and their associated shared resources are the companies that must succeed or die trying. Those companies are the VM companies.

I attended the Server Blade Summit last week. While the show focused more on the adoption of Virtual Machine technology (VMWare, Xen, etc) in the data center than on blade servers, some interesting trends regarding SAN attach rates for blade servers emerged.

In non-blade server world, the SAN attach rate (HBAs in servers) is somewhere around 15%. In blade servers, where direct attached storage is limited, the attach rate approaches 40%. And when VMs are added to the picture, the attach rate soars to near 80%.

Why is this? Virtual Machines are location independent. A VM can be suspended, put on the shelf and restored days or weeks later. And when a VM is restored, it can be restored on any processor blade in the complex. This requires that the storage associated with a VM be able to follow the VM to different processor blades. A perfect application for SANs and a driver behind the much higher attach rates for blade servers with VMs.

About the numbers- The 15% and 40% numbers come from conversations with IBM and HP. The 80% number came from a presentation I saw at the Server Blade Summit. If anyone saw the presentation and can properly attribute the numbers, please send me a note.

The advantage of FCoE is that is runs “over Ethernet” and not “over Fibre Channel”. Yet is connects easily to existing Fibre Channel SANs.

Why is the “over Ethernet” aspect important? “over Ethernet” allows servers to use a single networking technology, Ethernet, to access both network and storage assets. Using a single networking technology or fabric to access all of a server’s assets reduces equipment costs, reduces training costs and generally reduces complexity.

Why is the Fibre Channel aspect important? Fibre Channel SANs is where todays storage assets lay. Easy, efficient FC connectivity is a requirement for any new storage access protocol. FCoE, Fibre Channel transported over Ethernet, allows servers to connect easily to SANs.

The feature advantage for FCoE is less complexity.

As Marc Farley pointed out in his recent post (FCoE: Run Away, Its the Monster), new storage protocols are not new. Although FCoE was announced with great fanfare, the first draft of the specification is not expected to be ready until the second half of this year. The only real information that has been published are proposals made to the T11 committee; Convergence of FC and Ethernet: Host-side view, FCoE: Enabling Fibre Channel over Lossless Ethernet Fabrics, Call for participation on FC over ConvergedEnhancedEthernet

I like FCoE because the new name has two brand name technologies in its title- Fibre Channel and Ethernet :-)

Seriously, for any new protocol to survive in the mature (maturing?) storage environment, solutions featuring the new protocol must provide three things to its constituent consumers- 1) a feature advantage, 2) a cost advantage over the technology it is replacing and 3) it must integrate well with the existing infrastructure. I’ll address each one of these points individually.

A solution based on a new protocol must provide a feature advantage. A feature advantage is more than an incremental improvement, a feature advantage is not Fibre Channel moving from 4G to 8G. It must provide something that process technology along cannot provide. A feature advantage is an enabling feature- less complexity is a feature advantage. Ease of use is a feature advantage.

A solution based on a new protocol must provide a cost advantage over the technology that it is replacing. Cost is the total system cost, this includes management software, HBAs, switches, infrastructure and the trained team that implements and manages the solution everyday. Reusing existing infrastructure helps lower the the solution cost. A forklift upgrade does not.

A solution based on a new protocol must integrate easily with existing infrastructure. Greenfield deployments are rare. Most installations of a new technology will have to integrate with existing storage and storage management systems. You can’t force data to be migrated enmass to the new solution. A solution based on a new protocol must provide access to the legacy storage while providing the feature and cost benefits of new storage. A great storage solution will also extent those features and cost benefits to legacy storage.

FCoE, if done correctly, has a chance to meet the above criteria. If it does, it may not be just yet another storage protocol, it may be part of a real storage solution. I’ll address how I think FCoE can accomplish this in future posts.