Got a newsletter from AWS this morning, two great improvements that were holding me back from using it for production servers:
Nice - we can use this for more than just development/staging now. On the fly production environments without resorting to hackery.
Timely talk as we’re moving our dev and staging environment to be Xen-based. My major questions are scaling databases that weren’t setup for this infrastructure and production stability of the existing services. My raw notes:
- Virtualization - gave a quick overview of the state of virtualization.
-linux OS guys pushing back that hypervisors just putting overhead between kernel and user. “We can do buy you 95% of what Xen gives you right in the kernel, and do alot of cooler stuff.” Reminds me of what MS used to say about embedding IE into their OS.
-discussion revolved around this
-package management - virtualization makes it easy for you to get into a bunch of VM, then you enter into the normal package management problem. Versus baking it into the OS, and you’re just updating the one OS.
-these are the same problems that anyone who has ever scaled before has to face. The problems should be well understood.
-traditionally hypervisor (HV) considered more stable. But getting more and more complex so bugs in the HV are more and more likely.
-writing off hardware 5 yrs to move it off your books in US, 3-5 yrs in Canada depending on industry. VM allows you to skip the entire capital outlay.
-Baking it into the OS vs Vendors again - have to wait for the Vendors to release drivers, eg VMware only X number of SCSI drives.
-EC2 - how often do their IP leases last? Downtime two blocks of 7 hours in the last two years.
