VM export for EC2
aws.typepad.com
aws.typepad.com
I would love to hear more about the real barrier to migration: the cost and latency hit of cross-cloud network communication. Who cares if you can move your app server to VMWare, if it needs low-latency access to 10TB of data sitting in amazon-us-east?
As you say, the other difficulty is copying gigabytes of data across clouds. My colleague at Red Hat wrote snap[2] which can turn Linux guests into digests (basically a base distro name and list of packages and list of config changes), so this can be reduced massively. We're currently extending this work to support Windows and more Linux guest types.
AWS Direct Connect might help with that. Anyone have any experience with it?
"AWS Direct Connect makes it easy to establish a dedicated network connection from your premise to AWS. Using AWS Direct Connect, you can establish private connectivity between AWS and your datacenter, office, or colocation environment, which in many cases can reduce your network costs, increase bandwidth throughput, and provide a more consistent network experience than Internet-based connections."
Presumably you mean regions here? Within any one region, you can use an AMI in any of the availability zones.
At this point I have the entire process scripted, and it's still painful.
Edit: Oh, this is windows only, significantly less exciting. I wonder why linux isn't supported?
Linux runs on EC2 via Xen/PV, which is a kernel-hypervisor interface not provided by other (non-Xen) virtualization systems. Exporting linux AMIs would require reaching inside your filesystem and modifying your kernel, which is far outside of the scope of anything Amazon would want to do (both in terms of "really complicated" and "liable to accidentally break things"). The same applies to importing images, of course.
[I don't work for Amazon and I don't speak for Amazon. The above is speculation based on information published by Amazon about EC2.]
Amazon is unique in that they use PV-Grub everywhere to boot a customer filesystem kernel (newer behavior), whereas on other virtualization providers like Linode, the kernel is selected at domain creation time by the hypervisor, and loaded from the hypervisor's filesystem (not the customer). You can use PV-Grub but it's opt-in, not opt-out.
I'm betting you were running Xen/HVM then. (Most people do, these days.)
But you're right that Amazon isn't really interested in doing that work for you. Linux people who really need this are no doubt already doing it.
Last time I checked (admittedly a little while ago), one had to do API calls through some convoluted way.
I was tempted to script it.
Maybe Amzn is just not too skilled working with various bootloaders?