There had seemed to be a lull in the blogosphere and Twitterverse about Fibre Channel over Ethernet (FCoE) for a few weeks, and then, just as you would expect it would, traffic cranked back up as the fall tradeshow season roared into high gear. It has been a busy past few weeks trying to answer the classic five W’s (who, what, when, where and why) about FCoE and network convergence. As we all know, when we have a lack of real data, the speculation can sometimes get wild and wooly.
Here is Emulex’s take on the five W’s:
That is fairly simple, as at this point, the answer would seem to be everyone. Everyone of the major server, storage, operating system (OS), hypervisor and storage vendors have announced support of network convergence in 2009. This includes IBM, HP, Sun, Dell, EMC, Brocade, Cisco (Nexus and UCS), Emulex, QLogic, Intel, Broadcom, Juniper, VMware, Solaris, Microsoft, Red Hat, Novell and more.
This list marks a Who’s Who of the enterprise market. These organizations are creating the tools, systems, solutions and support required for IT managers to begin the process of testing vendor claims and putting pilots into the data center.
What is required for network convergence to become a reality? It is the confluence of virtualization, which drives I/O aggregation; Nehalem EX CPUs and chips, which set optimization for 10GbE bandwidths; blade servers leveraging 10GbE for mid-planes; and enhanced Ethernet, which addresses the needs of storage and high-performance computing (HPC) performance on a common physical layer. No one single thing of those listed above would move network convergence from concept to reality, but together, these items will, because it makes economic sense, leverages core IT capabilities and has the right problems to solve at the right time. These key problems are lowering the CAPEX cost of IT, increasing compute density to reduce OPEX (power, cooling, rack space, cabling) and simplifying IT to lower management overhead.
From the Emulex perspective, we have completed some of the major technology and ecosystem development steps over the past few years:
- 2008 – The year of driving standards and first-generation products, where, as an industry, we completed proof-of-concept products, performed initial technology demonstrations and built value propositions for our customers.
- 2009 – The year of OEM and ecosystem qualifications, as second-generation products have moved into the mainstream and OEMS are selecting their go-to-market partners and ecosystem partners, who will build turn-key solutions for IT. As this level of work is completed, IT managers can begin to look at implementing projects based on tools, equipment and services from their chosen IT vendors.
- 2010 – The year of validation, IT pilots and testing. IT managers will be able to put network convergence to the test and see if it really works for them. We obviously expect it will, after which it will become budgeted for the future.
- 2011-2014 – 2011 will be the first real year of deployment as IT managers replace equipment installed in the 2007-08 timeframe that has been fully amortized. This will mark the beginning of a three- to four-year transition that should set the stage for cloud computing and 40GbE network convergence.
Where will this occur first? The service providers (Web 2.0, financials, telco, IT services, X as a service) will be first, as they need to lower infrastructure costs while also increasing scalability and flexibility in their architectures. For many of these service providers, who provide back-end services related to networking, consolidating on 10GbE with enhanced Ethernet makes simple sense. They can provide easily provisioned bandwidth for each class of networking on a common infrastructure. This isn’t to say that all IT shops can’t benefit from this type of capability, but service providers drive direct savings, profits and revenue from this model and will be the most common early adopters of network convergence.
As we all know, IT vendors are quick to embrace new computing models, while IT professionals tend to be a bit more skeptical for many reasons. Network convergence is going to be driven first by 10GbE, and that transition is tied to the next generations of Nehalem EX servers running extensive server virtualization, which drive 10GbE to the LOM, 10GbE for virtual I/O and create blade servers to native 10GbE on the mid-plane. Clearly this transition has started this year, with HP Virtual Connect, IBM’s Virtual Fabric, EMC and NetApp announcing 10GbE iSCSI and FCoE support. As 10GbE becomes the basis of IT and the capabilities of enhanced Ethernet are added, network convergence will become a natural outcropping of the 10GbE transition. While network convergence is driven by the core of capitalism, it will lower costs, raise revenues, create efficient economies of scale and deliver a competitive advantage for those who implement it ahead of the market.