Spoken like a true software guy (no offense, I'm one too). Where we differ, though, is that I don't believe the ideal is a sea of unreliable bottom barrel x86 with janitors swapping out a box when the little blinky light turns red. No fun!

Also:
What all this pie in the sky open source hardware stuff seems to ignore is the most expensive piece in the whole enchilada: the CPU. And that certainly won't be 'open sourced' any time soon. "So what", you say. For that one piece, we'll just rely on Intel because that's what everyone uses anyway.
Well then, Intel would have more power than ever and would, no doubt, more easily be able to command more of the profits per box and use that to stomp out any competition, and then yeah! CPU monopoly!
But Intel wouldn't ever abuse that power, they're engineers, right?

I'm all for compatibility, but sometimes "fragmentation" is often the result of a *healthy* market. More unique designs to solve a problem is a good thing.

Snapshots are arguably outside of the scope of the hypervisor, but I can certainly see the value in your example.

Still, I would use LVM snapshots, or even better, netboot an OS image onto a ramdisk, and reboot it when you want to 'reset' it.
VMWare also includes I/O drivers in the hypervisor, which is another over-step of scope, IMO.

Some of the pain points of VMWare:

number of vCPUs, vMemory per VM
vSphere 5 now lets you have up to 32 vCPUs and 1024GB in a VM, which is good. vSphere4, which most people still, have is limited to only 8/256 per VM.

overhead:
VMWare takes a good 15-30%. Again the hypervisor in vShere 5 is a bit better performer.

stability:
I/O drivers are included in the hypervisor, which is a bit scary.

Pricing
The VMWare pricing model is overly confusing. Costs for added more vRAM to the pool? yuk.. vsphere5 makes this even worse.

I kinda see KVM as taking off here in the future. A lot of development is focusing that way..

Bar none, PowerVM still has VMWare beat in most areas that matter, but vSphere 5 is a step in the right direction.

If we're just talking x86, though, I keep hearing that KVM will be the top virtualization solution going forward.

Not to mention the record breaking droughts in Texas and the SW.

The data does suggest a global warming trend, but some weather shifts are normal. Just because floods wiped out *your* house this year doesn't mean the world is ending and it's all Fox News fault. In this case, the Pacific has been cooler than normal, which changes the jet streams. Which pushes moisture in the air in different ways. It's happened before and will happen again.
Just don't tell that to the dinosaurs.

There are Android-specific patches to the kernel, including an extensive security model, custom locking mechanism, and different frame buffer support among others. A lot of this code may have some ARM-only trickery. Add to that the library of redundant device drivers that phone companies write and discard (that may or may not work) and you have yourself a chunk of work there.
It doesn't help that Google has no interest whatsoever in getting their code merged into the Linux kernel properly. In fact, that's the main problem with Android. If their patches were merged and properly supported, device drivers would be better and it would be easier to do an x86 (or ppc or what have you) port. It's too bad since userspace is basically all java- the apps should just work on a new arch, but that benefit is torpedoed by Google's lack of follow-through in working with the community to get stuff merged.

In other news, Meego certainly seems doomed to hacker-land.

This isn't the first time I've heard about Google hiring in waves (they came into Austin a number of years ago, hired a bunch of folks, then closed the office, forcing people to either move or quit, now their back again).

So what are all these people working on?
There's the search engine, of course. And Gmail. And a slate of other hanger on type apps that no one I know uses. Google Books? anyone?
I guess they need folks for Android? I'm assuming that's the case. Now that they have a licensable tangible product to develop and support there's always work to be done and bugs to fix. OS's take a lot of work.

That's kind of a weird comparison, though. Power7 cores have 4 hw threads. Nehalem has 2 'hyper' threads.

Like any tool, you pick the right one for the job. Nahalem is quite fast on a single thread, but if you have a web server processing boat loads of transactions/second, you may look towards a tool that is fast on many theads and can churn through many transactions concurrently.

They did *buy* the original Android. Are they hamstrung by the decisions that were made in that early version? It would just be good to know, I think.

Maybe they should change their moto from "Don't be evil" to "Don't be a dick".

That may be true, but that's what they said when those initial USB jailbreaks came out. That method juked the PS3 into going into debug mode. Maybe this joker found another way to do that?