Is the idea of connecting Fibre Channel devices directly with Ethernet networks a bad idea? Is creating a new standard to help it happen a way to keep customers happy by helping them keep their expensive storage systems up to date? Or is it a cynical attempt by Fibre Channel vendors to stave off more efficient, less expensive IP-based storage to preserve their own market and keep the money flowing in?
Depends on who you talk to, apparently.
In our story last week, a group of storage manufacturers made the case that the Fibre Channel over Ethernet (FCoE) specification they submitted to the American National Standards Institute would help customers and the channel make the transition to IP. ISCSI manufacturers made the opposite case.
It's not a small issue to the people who buy or sell the systems, so we gave each side a little more space to make their arguments.
First up is Doug Ingraham, Brocade's senior director of product management, who spoke by phone to SearchStorageChannel.com's senior news writer Nicole Lewis.
Click here to read EqualLogic's opposing view from John Joseph, vice president of marketing for the enterprise-class iSCSI vendor.
Nicole Lewis:There has been some speculation that larger vendors don't want to support iSCSI's advancement because FC is a more expensive infrastructure and they'll lose profit margins if iSCSI advances. Do you think large vendors see the potential growth of iSCSI as a threat?
Doug Ingraham: Data centers worldwide have an enormous investment in Fibre Channel SAN infrastructure. That global installed base has thoroughly demonstrated its end-user value in providing high performance, high availability, data protection and economies through consolidation of enterprise storage assets.
iSCSI has not demonstrated the performance and proven track record of manageability for mission-critical data center operations, but has proven to be an excellent technology for integrating low-cost second-tier servers and applications into data center SAN best practices.
iSCSI is also suitable for moderate-performance departmental applications, but implementation for first-tier applications really requires 10 Gbps and the higher costs associated with 10Gbps undermines iSCSI's value proposition as an economical interconnect.
FCoE will first be deployed by enterprise customers who will combine their storage, high performance computing and data networks on a combined infrastructure. This will eliminate the number of different adapters, NICs and HBAs customers need in their enterprise servers. This will run over a lossless Ethernet, which will be well suited to data center distances and environments. Because this is first going to appear for enterprise customers you see the larger vendors being the first to promote the technology. All of those same vendors also have iSCSI products to address different storage networking market segments.
Fibre Channel, FCoE and iSCSI are complementary technologies and will likely coexist in data centers for the foreseeable future. Even large vendors have midtier product offerings, and it's difficult to imagine that any vendor would leave a market opportunity untouched based on which protocol is used. Brocade, for example, has both enterprise-class Fibre Channel solutions and iSCSI products, and is also actively promoting development of the FCoE standard.
Lewis: How will the FCoE specification affect iSCSI's growth?
Ingraham: Currently, FCoE and iSCSI are playing to different market segments and application areas. One of the main attractions of iSCSI is its ability to run over commodity 1 GigE infrastructures and standard 1 Gbps NICs. The protocol itself is free. It lacks the performance and management capabilities of Fibre Channel, however, and so does not compete in multigigabit data center applications. Rather than try to get iSCSI to walk, talk and act like Fibre Channel, customers can celebrate its cheapness and use liberally for servers and applications that do not require high performance and scalability.
FCoE's main attraction is its contribution to the general trend towards a unified transport infrastructure (a.k.a. unified wire or data center Ethernet) that offers very high performance and flexibility to mix different protocols and applications on the same delivery system.
FCoE is not intended for low or moderate performance applications (on which bandwidth would be squandered) but for the next generation data center infrastructure. Compared to iSCSI, FCoE retains Fibre Channel's low protocol overhead and efficiencies but it is unlikely to slow future growth of iSCSI for second-tier applications.
Lewis: If it does affect iSCSI growth, then what will this mean for your company and particularly for your VARs?
Ingraham: Although Brocade has historical roots in Fibre Channel, we are protocol agnostic and are aligning a diversity of products to suit different customer needs. This is demonstrated by Brocade's further development of Fibre Channel for both data center and mid-range applications, FCoE for next generation enterprise SANs, file area network solutions for IP and WAN environments, iSCSI gateway products, as well as professional services across all storage technologies. Brocade's channels and VARs therefore have more market opportunities to meet a wide range of storage needs, including current and future iSCSI deployments.
Lewis: Do you think this specification will lower the cost of FC to the point where it will become more attractive to SMBs, which for the most part is where iSCSI is attractive?
Ingraham: The potential cost benefit of FCoE is primarily in facilitating a common transport for multiple protocols and applications in the data center. A 10 Gbps data center switch with FCoE storage services will still be more expensive than today's commodity 1 GigE switch. Small businesses on limited budgets can still enjoy the benefits of shared storage using iSCSI, standard NICs and Ethernet switches, but this is a different market area compared to the data center applications FCoE is targeting.
Aside from FCoE, Fibre Channel products in general have undergone steady price reductions and have become affordable for SMBs. An SMB that needs the 4 Gbps performance of Fibre Channel, for example, can deploy an economical SAN solution using Brocade's 200E fabric switch. iSCSI at 1 Gbps cannot provide a comparable solution, and at 10 Gbps would be too expensive.
Let us know what you think about this story; email: Nicole Lewis, Senior News Writer.