Selecting Persistent Cloud&Heat StorageΒΆ

This section contains information which is very important to consider when using Cloud Compute. When data is written to the file system on a running VM, it is accessible as long as the VM is running. But once the VM has crashed or has been terminated (as opposed to shut down), the data is lost. This kind of storage is called ephemeral storage.

There are several ways to avoid this problem. For example, one can make a habit of taking periodic snapshots of the running VM. These snapshots can then be used as a backup or checkpoint. For example, if you are installing or developing something that you know is risky, you can first take a snapshot and, if needed, do a rollback by terminating the instance and launching the snapshot. This is however only an alternative if you can allow the instance to be offline until the snapshot has been relaunched. In many situations this is not an option.

Another way of solving the problem is to attach persistent block storage to the instance once it has been launched. Cloud&Heat calls the pieces of persistent block storage volumes, and the data on them is replicated three times. Volumes can be thought of as external hard drives that you can attach, detach and reattach to your VMs. You can attach a volume to many different instances (but only one at a time), and you can simultaneously attach several volumes to the same instance. See Cloud Block Storage Tutorial for further information.

Object storage is another option. With Cloud&Heat Cloud Object Storage (based on OpenStack’s Swift project), you store files as objects in so-called containers. They are stored in a flat hierarchy together with a unique identifier which is used to find the object. Because the way the metadata and physical storage are managed, this kind of storage is extremely scalable, meaning that you can increase the size of each container seemingly endlessly. Another advantage of object storage is the convenience with which you can access your files. You can access each file using a browser by pointing it to a unique URL, or you can treat the object store as a file system with the help of third party software such as s3ql.

There are also programs, for example Cyberduck, which provide graphical user interfaces which allow you to easily drag and drop files into containers. Or you can use the containers to save automatic backups of your system, using, for example, Duplicity.

Refer to Cloud Object Storage Tutorial for a general introduction, and to Cyberduck Tutorial and Duplicity Tutorial for specific information regarding these solutions.