Using commercial cloud providers has many advantages: cheap, efficient, allows you to work in multiple locations. Just remember that unless you have a specific service level agreement (SLA) in place, there isn’t necessarily a guarantee that your data will be retrievable.
A slideshow by Skyhigh Networks points out, for instance, that even if you are using a dedicated instance of EC2 on Amazon Web Services, there’s a risk your data might disappear during a compulsory reboot. The user agreement specifically states that instances can be rebooted as often as every three months, and that any data lost during a reboot is not Amazon’s responsibility.
If you have replica instances and other backup elements in place, this is unlikely to be a problem (and as our first commenter points out, EC2 is primarily for execution instances, no permanent storage). However, it serves as a reminder that conditions for off-site providers need to be checked carefully, to make sure they do match your own business risk appetite.
6 Sneaky Terms That Tech Companies Force You To Agree To [Business Insider]
Comments
2 responses to “Reminder: Cloud Data Storage Is Far From Guaranteed”
EC2 is not for permanent data storage, never has been, for that you should be using s3 and glacier. This ‘article’ is like pointing out the using a car as a submarine is not a good idea
Yeah I was about to mention the same thing… Spastic fucking article, anyone running instances on EC2 would be extremely likely to have a copy of their own data.. The most they’d be likely to loose is their finished result.
You two are missing the point. The article is talking about the policy for *running instances*. It is very important to note that there is no host resiliency on EC2. So once a quarter, your host will get rebooted and your instance will take an ungraceful “power off” event. You can architect around this, but you absolutely will use transactions if anything hasn’t yet been committed to the EBS volume (in memory, in cache, pending, etc). EC2 paired with EBS volumes absolutely *is* for permanent storage, but whether you are using EBS, S3 or some object store that isn’t even part of AWS, you are still at a potential risk to lose data because no EC2 host is highly available and they *do* get rebooted periodically for maintenance with no warning.