Excerpts from Paul Graydon's message of Fri Jan 06 01:12:03 +0100 2012: > > There has been a bunch of talk recently around blogs that is confusing > me a little bit, the latest of which seems to be a debate about whether > you should use EBS rather than local instance storage for / or not.
You have to keep in mind that amazon gives no guarantee that EC2 nodes will work reliably or will have such or such % of availability. EC2 nodes can crash or fail, it's a fact. And it actually really happens sometimes. Using EBS volumes for your instances is quite similar to real hardware: when a node dies you can "unplug" the "hard drive", plug it in another node, and restart the service there. This means you'll need to check the monitoring and have a script or a monkey actually do the job in the case of such an event. Using ephemeral volumes for your instances, you have to consider that when an instance crashes, everything it holds is gone. This forces you to completely automatize provisioning, configuration and application deployment. It forces you to make everything redundant and stateless by default. It forces you to think of the infrastructure as a whole and not focus on individual nodes. This is more work, but it allows cool things such as auto-scaling, duplicating your infrastructure for tests/dev/migration purposes or migrating all your infrastructure to rackspace within an hour. So using EBS-boot is definitely easier if you're new to AWS. But instance-store enforces good practices and is IMHO the way to go once you start having more than a couple of nodes in your EC2-infrastructure. > Anyone got any particular observations, hints or tips about what I > should expect on AWS? Do not rely on fast disk I/O... Have fun ! Cheers, Marc _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
