Friday, August 3, 2012

Cloud Storage is more than just Backup

Without a doubt, Cloud Storage is front-of-mind for Storage Professionals, System Architects and, really, anyone working in the IT field.  The extreme scalability and pay-for-what-you-use economy makes Cloud Storage a very attractive prospect and an easy way to extend an existing storage archive infrastructure.   While Cloud Storage is a great off-site backup for your data, treating Cloud Storage as just a "tape silo in the sky" misses out on a lot of power and capability.

The term "Cloud Storage" is bandied about a great deal has come to encompass any kind of Internet-accessible storage which is paid for on a per GB per month basis (or sometimes, even free).  This definition covers file sharing and archival systems like Dropbox, Google Drive, and iCloud, but it also covers service offerings like Rackspace Cloud Files, HP Cloud Storage, and Amazon S3.  In reality, these services are all built on a new storage paradigm called "object stores."

Understanding object storage is pretty simple; it is a system which lets you push in an object (a file) and returns a unique key for that object.  That key can then be used to modify that object, read it, or delete it.  Object stores also let you set metadata on your objects, allowing you to expose useful information about an object without having to actually download and read the object.  Simple, but this simplicity allows object stores to scale to amazing sizes and give consistent performance -- regardless of the number of users.

The flat key-value way of storing objects may sound limiting, but if you think about it, its actually very flexible when metadata is used.  Think about trying to organize your photo collection -- on a file system, you create a folder hierarchy based on, say, the month and year the photo was taken.  But, if you want to later group your photos by who is in it, with a file system you have to make copies, or move to a new folder hierarchy.  In an object store, all your images are in the same level, but metadata for date taken and subject is set on each image, allowing you to view your photos by date, subject, or any other tag set on the photo.  And it will do this whether you have hundreds, thousands or millions of photos.

The interface into object stores is an HTTP-based REST API.  While this might be a challenge for some, it serves as a boon to anyone with a service trying to use that data.  Take, for example, the HP Cloud CDN - your objects are easy to deliver through the CDN because all the requests are HTTP requests on the objects themselves, instead of a file buried somewhere in a hierarchy, and are delivered directly out of the object store, rather than having to be pre-staged on a web server.

Object Stores are powerful and flexible, and is how services like Dropbox, Pandora, Instagram, Twitter, and GMail (to name a few) are able to store, manage and deliver their content to their users.  Sure, its inexpensive, reliable, offsite storage, and makes a great addition to an IT disaster recovery plan.  However, object storage is capable of so much more.  To store, search, and process the never-ending volumes of data we are generating, businesses should think beyond just Backup.


No comments:

Post a Comment