I have been watching (and occasionally poking my 2 cents in) a few interesting discussions on Twitter lately about the proprietary scope of various vendor network fabrics, like Brocade's Virtual Cluster Switch, Cisco's FabricPath, HP's Intelligent Resilient Framework and Juniper's QFabric. I am interested in the topic because of the level of commitment an enterprise has to make when choosing and deploying one of the fabrics. It's a big commitment: Rip and replace is really an expensive option once you choose one vendor over another, and all of these fabrics involve ripping and replacing your existing gear. In nearly all cases, the replacement of data center networking is proprietary. Yet when we ask IT about proprietary vs. standards-based purchases, the majority of responses indicate that standards are nearly always preferred. So why are the network fabrics proprietary? There is a disconnect between what we are hearing from IT and what vendors are offering, and it seems to me that vendors are collectively shooting themselves in their feet.
Year over year, our research shows a roughly 50-50 split between IT people who want single vendor suites and those who want multivendor, a.k.a. best of breed. This is consistent across verticals, company size, revenue and location.
In our most recent survey of 444 business technology professionals involved with LAN equipment, we asked a different question, trying to get a better handle on respondents' preference for standards or proprietary products and when they might use either (this is from a larger data set I am analyzing). The data shows that most respondents, 67%, think proprietary products are to be avoided if possible, and a large minority, 32%, think that some proprietary products are acceptable. Only 1% think proprietary products and features have better integration and interoperation features than standards-based products.
In our 2010 IT Pro: Data Center Networking Vendor Evaluation Survey[subscription required], (we are updating that survey even as I write), multipath Ethernet, which underpins Ethernet fabrics, was ranked 13th out of 15 must-have features. In that same survey, "Proprietary features in advance of standards" was 15th out of 15. That's bottom of the heap, just to be clear.
There is a disconnect between what our surveys show IT wants (standards-based products) and what vendors are delivering (proprietary products). What to choose is a big decision because we are talking about an investment of time, money and expertise to roll out an Ethernet fabric. What you buy today will set the direction of your network for the next several years--most likely several product cycles, at the very least. Companies, at least none that I have talked to, don't re-evaluate vendors and switch every three to five years unless there is a very compelling reason--like betting on the wrong technology, such as Token Ring or ATM. Companies tend to re-invest within a single vendor or product family. The total investment spans a long period due to factors such as inertia, the cost to switch, avoiding disruption or pressing dependencies on the status quo by other critical systems.
Standards-based products make swapping easier because you can take another product that supports the standards, configure it and, in theory, swap it with nary a hiccup. I know it's not that easy, but it's far, far easier than swapping proprietary stuff. You can swap a Cisco router with a Juniper router and it all just works because the entire stack is based on standards. But just because IT can swap gear doesn't mean that they actually do swap gear. At least, they don't swap gear on a whim, but there is a strong practical and psychological comfort knowing that they could and that they aren't getting locked in to a proprietary product.
When I ask vendor spokespeople why their companies are creating proprietary products when there is clearly standards work going on, the usual reason I hear is customer demand. "Customers want [the new technology] now," the vendors say, and usually follow up (off the record) stating that they'd love to have standards but the standards bodies are slow to deliver standards documents and our customers can't wait. See, it's the customers' fault that vendors make proprietary products. (OK--that last bit was just me being ridiculous.)
Fibre Channel over Ethernet (FCoE), arguably the primary driver behind multipath Ethernet, may or may not be dead, although Doug Gourlay, VP of marketing at Arista, a former Cisco employee and generally smart guy thinks it is but we don't know it yet. I think FCoE is either going to be on a slow adoption curve like VoIP was or, like VDI, there will be a few deployments and lots of tire kicking that turn to naught.
However, what I do see is this: Most enterprises, large and small, are not going to jump into a big investment in what they perceive to be proprietary products. It's one of the reasons that hurt VoIP adoption--the requirement to purchase within a single product line or partnership program lest IT take in the responsibility of making handsets work with PBXes. (Don't give me the "SIP is standard" line. Try to get phones and handsets to work reliably outside of hardware compatibility lists). It's the reason why 802.11 Wi-Fi adoption was slow in both consumer and enterprise IT until the Wi-Fi Alliance formed and hammered out an interoperability testing and logo program ensuring that any NIC and AP would work together reliably. I don't think I am being hyperbolic in saying if it weren't for the Wi-Fi Alliance, 802.11 Wi-Fi products would be crippled today.
Enterprise IT wants to know that their products are going to work with the rest of their IT systems, and unless there is a really compelling reason to adopt proprietary products, they probably won't. It's not that Brocade, Cisco, HP or Juniper are likely to close up and leave customers hanging. It's that once enterprise IT commits to a product path, they are committing for years and creating dependencies on the network. The don't want to be hobbled by proprietary products. Fair or not, IT has a long memory, and no one wants to let a vendor dictate IT direction.
I think the stall in adopting multipath Ethernet has as much to do with IT hesitancy to commit to proprietary products as it does with factors such as price and learning curve. I think if the vendors--all of them, not just the ones I mentioned--would sit down, work out the standards, move them along through the standards bodies and commit to interoperable products, the whole market would grow, which is good for everyone--vendors and customers alike.