Musings of a West Texas Business Venture

Amazon.com's EC2, Part Three

EC2's weakest "point" is that data inside of the virtual machine only lives as long as the machine is powered off. When the machine turns off, all of the data is lost, unless it has also been saved somewhere else. One way to store data off of the instance is to use Amazon's S3 service, which works fine for storing "regular" files, but does not work as well with remote storage of real time information, it could take more than 10seconds for file changes to appear on S3. For many applications, for instance database storage, S3's service would be extremely cumbersome.

A few third-party tools have appeared to try and help solve the remote file storage problem, the most prominent of which is S3InfiniDisk. S3InfiDisk is a filesystem, based on the FUSE project, and includes caching of data, and multi-write and multi-read access to S3's storage buckets. I have not yet tried out S3InfiniDisk, but its creators claim that it is fast enough to use beneath a modern database, like PostgreSQL or MySQL.

We will likely continue to use, and evaluate, EC2 for our infrastructure needs, both internally, and for our clients.

posted by Anthony Presley @ 11:04 AM