One of the most common words to frequent conversations in the computer industry today is ?cloud.? Seemingly, this single word would represent a singular approach, but that is far from the truth. There are many types of applications/uses for the cloud. There has been a major push to take software applications to the cloud (Software as a Service/SaaS). An even broader approach takes the compute process as well as applications to the cloud (cloud compute).
One of the most popular uses for the cloud is transcoding and distribution ? the ability to convert a single media file from its source format into multiple formats which allow the file to be played back on various devices such as tablets, smartphones, PCs, etc. ? and then distribute it to users. Backing up files to the cloud has also become a popular process, especially for small data sets such as on smartphones or personal computers.
A relatively new approach to cloud use involves backing up large data sets to the cloud. This holds great appeal for organizations wanting to move data offsite for disaster recovery. Cloud backup itself is not new. Smart phones and personal computers have used cloud backup services for many years to backup small amounts of data. What is new is attempting to move hundreds of terabytes, or even petabytes, to the cloud.
The bandwidth to move large data sets, and more importantly restore large data sets, is both cost and performance prohibitive. In moving data to the cloud, the upload or backup process can be accomplished by ?trickle feeds,? slowly moving small amounts of data until all data has been moved.
If this approach is used for disaster recovery, a ?trickle restoration? is insufficient when some form of disaster has wiped out an entire server, server groups, or an entire data center. It could take a week or more to restore 100TB over a relatively fast internet connection. Amazon offers their Snowball and Snowmobile services for bulk upload, and this approach may have merit for some customers, but these products don?t significantly change the economics of cloud backup.
This paper is designed to examine the true cost of storing data in the AWS Glacier Cloud versus in an Iron Mountain Storage Facility. To source data for this paper, we used Spectra Logic as a ?real-world? example (Spectra uses both Amazon and Iron Mountain), and obtained actual quotes from both vendors.
To allow users to make calculations and decisions based on their own data sets, we start by looking at the cost of storing 6TB of data (a single LTO-7 tape) using Amazon Glacier versus Iron Mountain.