Real Differentiation: A Rare Commodity
I hear it so much from customers, bloggers, blog critics, the media, that servers these days are a "commodity". I just can't agree. For starters, let's look at the definition:
commodity. (n.d.). Dictionary.com Unabridged (v 1.1). Retrieved September 27, 2007, from Dictionary.com website: http://dictionary.reference.com/browse/commodity
| 1. | an article of trade or commerce, esp. a product as distinguished from a service. |
| 2. | something of use, advantage, or value. |
| 3. | Stock Exchange. any unprocessed or partially processed good, as grain, fruits, and vegetables, or precious metals. |
Isn't it fascinating how these definitions are completely opposite from the colloquial usage? A product distinguished from a service. Something of use, advantage or value. Hardly the terms commonly associated with an ocean of rip-and-replace servers.
Not too long ago the idea of lumping pork bellies, copper and enterprise servers into the same economic bucket would be an unconvincing proposition. Now, it's with almost universal relish that people cast certain things down as-- dare we say-- a commodity. The cynics say that trying to differentiate is a losing battle. The curmudgeons say trying to differentiate is stupidity. I say that nary a server has ever been sold that was not, in the eyes of its purchaser, different from the competition. In fact, by definition if you have chosen a vendor then you have applied some set of metrics that differentiates them.
Let's say, for a moment, that we indulge the colloquial and sadistically pejorative meaning of commodity as used in the IT world. Is it really the case that servers are commodities? Do IT organizations extract absolutely zero value from choosing a vendor?
This question can be answered on two fronts. First, we can try to address this by doing all sorts of TCO calculations or comparing speeds and feeds or timing delivery speeds. I prefer to take a more unbiased and demonstrative approach: show me the evidence. Show me the customer that truly benefits from getting every server it buys from the lowest bidder every time, without any regard to how many different vendors might ultimately be represented. You won't find one. Not one that has its costs under any kind of control, anyway, which is the whole point of commodity markets.
If I went to a small, medium or large Dell customer and I told them that, tomorrow morning, all their servers would be replaced with HP's, what would people start to think about? Would the sys admins wonder if the system consoles would be the same or if it would drop in to their management infrastructure? Would the developers wonder if the performance characteristics would be the same and cause an entire regression testing cycle? Would the IT Ops director wonder if the service levels would be the same and how that would affect internal and external service level agreements? Would the CFO wonder how all these changes will manifest themselves on the balance sheet?
The answer is that no one in an IT organization would expect a vendor-for-vendor replacement to be zero-sum. That's not to say that things couldn't be better after the switch, but at the very least they will be different. The difference comes from vendor characteristics and product characteristics.
If you are the Hormel Corporation, a pork belly is a commodity (in the colloquial sense) if it consistently meets its expectations: 12 inches square, exactly 5 pounds and 1 inch thick. If I told Hormel that half the bellies would be 2 inches thick and 6 inches square they might surely have a preference for their source. In addition, the pork belly requires no "service" from its vendor. No one needs to install a pork belly or have a maintenance contract on it. (I bet there's a sentence no one has ever read before.)
Conversely, if you are a consumer of space at a co-location facility, you might feel very differently if I told you that you would need 20 racks instead of 10 because those servers you thought were a "commodity" do actually vary considerably. In addition, the service required on 10 racks of machines will surely be different from that required of 20.
The truth is, it's not the server or server vendor that is the commodity, it is the architecture. The X86/X64 architecture is a commodity-- itself totally void of bias, added value or, as a self contained unit, differentiation. The same is true for SPARC, which is open-sourced. In this sense, being a commodity is not a market limiter but an avenue for terrific market opportunity.
If vendors consume architectures as a wholesale commodity then the value they add is the implementation, packaging, delivery model, support model, operating system support, etc., etc., etc. All of these take an unfinished architecture and make it in to a useful product. A product that is not, by the way, divorced from its services.
That is why, despite the howling of all the cynics and curmudgeons, Sun's market share keeps growing. Indeed every vendor recognizes that having no differentiation is a heck of a lousy way to sell product. Sun recognizes that differentiation in the form of truly innovative products that reduce opex is the way to provide the best value to its customers.
That is a rare commodity.
