Modified FCoE Adoption Curve

So, David Vellante created an excellent FCoE adoption curve posted on Wikibon today. When the conversation about “how quickly” or “how much” is applied to FCoE, whether it be hype, fact, or perception, I believe that David’s curve is an excellent broad based timeline that can be seen as a great metaphor. However, I think there is a major shift in the curve’s shape that should be included. Sadly, that shape is a slight downturn before we hit the full upswing.

Adoption of any new technology must take into consideration the consequences of those change. This becomes even more important when something as public as FCoE emerges into the marketplace. That is to say, people will talk.

For the record, I like (and agree with) David’s ROI graph in principle. There is, however, a momentary hiccup in the graph that he omits out of expediency that should be re-inserted. I believe that the more likely adoption curve will look something like this:

Adoption Curve

Adoption and Return on Technology will hiccup before finding its momentum.

While the general direction of the curve follows David’s, there are a couple of elements that have some tremendous impact on the adoption of technology.

I agree that the tepid growth of FCoE as a Data Center (DC) technology (1) is marked by typical early field trials (EFTs) and pilot tests. This is also the time when the most buzz and “hype” about FCoE will have its greatest push, the most controversy, and the greatest number of questions. In other words, right now.

When vendors begin releasing ready-for-prime-time equipment, as noted in by the star in this next graph, Customers begin to examine the real-world possibilities and begin making plans. This is the true first-generation “production ready” cards and switches.

Adoption begins to pick up as vendors release 1st Gen "Prime Time" products.

In reality, this is more of a phase than a moment in time. For instance, QLogic released its single-chip CNA in FYQ2 2009, Brocade FYQ3 2009, and Emulex recently announced theirs, though as of this writing has not been generally available in public release. However, customers are beginning to take notice of these events as not just theoretical but opportunities for very real implementations.

There will come a point however when the natural limitations of adoption will start to have an effect, and that adoption will slow (as well as the return on investment) (2). With FCoE, there are two causes for this, one technological and the other cultural.

First, in very real terms FCoE is – as currently standardized – useful for limited deployments. This is okay, though, because FCoE is a “bolt-on” technology, which allows you to evolve the data center, as opposed to rip-and-replace components. Nevertheless, as people begin to realize that there is no multihop support for FCoE yet, they may hold off on the budgetary resources for FCoE adoption.

For what it’s worth, I believe this is would be a significant strategic error, for reasons I’ll get to in a second. However it would be foolish to ignore that the adoption rate for FCoE will slow down as these standards (among others) are yet to be ratified. It may not be the best policy, but it will happen nevertheless.

The second, more cultural reason, is that the natural silos that data center managers enjoy in terms of controlling and separating their respective bailiwicks will need to be torn down. FCoE, unlike some of the ad-hoc approaches that have spawned data centers that look more like Lego™ construction sets, requires a great deal more personnel planning than before.

The peak of the curve (2) is when individual companies begin to realize that they have not yet asked the question, “Who’s in charge?”

Should it be the network guys? The storage guys? The server/application guys? In large data center environments these are usually isolated and incommunicative groups. If your organization does not sound like this, congratulations: write a book. 🙂 You are the exception that proves the rule.

It’s during this part of the adoption that discouragement sets in. The questions begin: “Why did we do this?” “What were we thinking?”

Then the accusations: “It was all just hype,” “We were sold a bill of goods.”

This is the part where people will simply give up, say that the technology isn’t ready for prime time and go back to doing inefficient and ultimately costly behaviors that were the reasons why they were looking for something new in the first place. In other words, they will voluntarily set themselves back farther than when they started.

When that happens – and it will – the return on that investment will slide. There’s a regrouping, a need to get the footing again, and let go of assumptions about the technology and the personnel that will be holding companies back.

What assumptions? Well, to start with, FCoE is not a panacea and your data center guys will not “just sort it out amongst themselves”. Both are myths, not true, falsehoods, fairy tales, and no, Virginia, there is no such thing as an FCoE-fairy.

Reformulation of expectations will finally push FCoE forward.

That return on investment, and adoption rate, will begin to pick up at the convergence (pun intended) of three distinct events occur (3). It is within this range of activity that David’s Breakthrough happens.

First, the T11 group will ratify the remaining outstanding elements of multi-hop FCoE traffic, including (but not limited to), congestion notification and Enhanced Transmission Service. Once these technical specs are solidified into a usable format then vendors (who are already testing various strategies) will be able to provide rev2 products for the marketplace. These new products will in all probability remain backwards-compatible with rev1 and continue the “evolutionary” approach.

Second, the process of injecting a unified, converged system into a traditionally isolated cultural environment will become codified. Granted, us DC people are all geeks, but we’re not all the same geeks. Wise is the CIO or CTO who recognizes this fact in advance, and at this point along the adoption curve there will be appropriate methodologies created to engender a cooperative working environment before political and cultural damage is done that would sabotage the project.

Third, as the economy begins is slow, sluggish climb out of the depths of uncertainty, aging equipment will be due for refreshing. The companies that have begun the process of EFTs with FCoE will have the greatest head-start on adoption, having learned the lessons with small testbeds and learning environments while still gaining the reward of lower OpEx. Core switches will be released and a strong push to Native FCoE environments will become common.

For this reason I highly recommend avoiding a complete wait-and-see approach: either it is or it’s not understood that convergence is going to drive data center development over the next ten years. It’s easier to have a culture adapt along with the technology than to attempt to turn it on a dime and simply hope for a smooth change.

Hope is not a strategy.

Only when those expectations are met with reality, both internal and external to the organization, will the strong push and momentum for FCoE begin a steady climb (4). The technology will have reached enough of a critical mass to provide multiple case studies for deployment in a variety of data center environments. The FCoE vendor ecology will self-organize to provide standards-based approaches that encompass different types of traffic at different QoS.

Internally, Data Center culture will begin to morph to include at least a couple of different models of organization and management in order to plan and phase in FCoE technology as a process.

Then, at the time David calls “mopping up time,” we’ll do it all over again with 40GbE and 100GbE, and then complain that someone should have thought of how to make the process easier a long time ago.

You can subscribe to this blog to get notifications of future articles in the column on the right. You can also follow me on Twitter: @jmichelmetz