The Race to the Bottom — of the Software Stack

In politics, it’s said, your enemy’s enemy is your friend. That formulation is true in technology as well, as evinced by the news that Dell is working with VMware to create a server with virtualization embedded in its firmware.

What motivated both parties to participate in this? In one way, you might think this kind of deal makes no sense for Dell — after all, if it delivers a way to make using virtualization easier, won’t the proliferation of virtual machines reduce demand for physical servers — you know, the kind Dell sells?

That’s true. It seems a little unappetizing. But look at the current situation from Dell’s perspective. It does all the hard work of building demand with an army of salespeople. It does all the work of building the machines themselves — dealing with the supply chain, distribution, and so on. And then it loads Microsoft Windows on top of the box. And Microsoft makes most of the margin on the box. From Dell’s perspective, it’s doing most of the work and getting little of the profit.

So, an alternative that lets it avoid the Microsoft tax looks pretty appealing, one would guess. And, compared with dealing with the legendarily tough Microsoft sales team, VMware must look like a much friendlier partner.

From VMware’s perspective, this is a good move as well. While the company rules the virtualization roost today, Microsoft is coming right after it in the upcoming Server 2008 release. Its hypervisor, codenamed Viridian, trods the same path as many previous Microsoft strategies: let someone else develop a market and then come in with a cheap/free alternative that, by virtue of being bundled with the operating system, makes the existing option look less interesting. I mean, if you’re getting functionality for free, why would you go out of your way to buy one? So VMware has looked at its predecessors like Netscape and doesn’t like how the story plays out.

This is just one more move in the disintegration of the once-mighty Wintel monopoly, a razing played by both Intel and Microsoft for their own self-interest. The rise of Linux, aided by Intel, and the rise of AMD, aided by Microsoft, have both been motivated by the desire to reduce the power of the other party and increase leverage in the relationship.

These moves, and the move by

VMware and Dell to offer yet another option, are all driven by the same impulse: to more tightly bind the primary relationship with the customer. Whatever company provides the strategic part of the software stack, the primary piece that end users think of as the key purchase, reaps the lion’s share of the profit available from the transaction. And this is always the rule in markets, so you can imagine how galling it is to Dell to do all the work to create the sale and then give up most of the profit to Microsoft.

And so this announcement from VMware and Dell indicates their desire to race to the botton of the software stack and realize the profit potential available there.

It’s a brilliant maneuver. It lets Dell break out of the current game and improve its margins. It lets VMware sidestep its ordained role as irrelevant pioneer.

So, the motives of the participants are clear. But what will they deliver?

While the details are not yet available, some hints are dropped in the article linked to above.

The new server line, called VESO, will feature dual quad-core AMD Opteron chips (I guess if you’re changing the rules of the game, you might as well break them all), be able to handle 256 gigs of memory, critical to virtualization, where memory is usually the bottleneck to scalability, and contain an embedded VMware hypervisor.

When I first heard about this a few months ago, my immediate thought was that there was no way to accomplish this, technically speaking. The size of the VMware hypervisor would preclude it being embedded in anything like the amount of firmware memory available on motherboards.

But in discussing it with my friend Alan Horn, he pointed out the Linux kernel itself only runs a couple of megabytes (and is often stripped far below that for embedded systems), so VMware’s task might not be impossible.

If I had to guess, I would wager that the embedded hypervisor will deliver virtualization hosting with rudimentary management capabilities with a way to load additional disk-based software providing more sophisticated management, along with a strong push to move to VI3, WMware’s server pooling product.

This functionality arrangement lets Dell avoid delivering Windows servers — now it delivers Dell Servers, and lets the customer decide what guest OSs to install.

And, with the coming move to applications being delivered as software appliances, this could move Microsoft from a dominant position to a supporting position, just where Dell would like it to be.

This functionality arrangement would also allow VMware to trump Microsoft’s bundling of virtualization into the operating system by bundling virtualization into the BIOS of the system. Microsoft would be hoisted on the petard of its own bundling strategy! This arrangement of functionality would also provide VMware with upsell capability, thereby avoiding the joyless anonymity of being a mere OEM supplier.

I’m really looking forward to the release of this new server line from Dell. And, unless Dell was able to negotiate extended exclusivity, a la AT&T with the iPhone, we will see similar offerings from the other hardware manufacturers. Because everyone ‘s looking for enemies of their current “friends.”