Difference between backup and archive

[mgarczynski]

I am new to TSM application and am running version 5.5. I want to understand the differences between backup and archive as the terms are different from my previous backup solution.
I believe I have backups working properly and the data retention appears to be working fine.

However, my historical data is not. With my previous backup solution, I was able to have several rotation/retention schemes running. Those schemes were daily which expired media after 6 days, weekly which expired media after 3 weeks and monthly which expired media after 60 months.

Is a similar scenario possible with TSM? If so, do I use backup or archive? Are there best practices for TSM?

 

[dms19]

In a nutshell:

Backup

The backup process copies data from client node to TSM server storage. The server retains versions of a file according to policy, and replaces older versions of the file with newer versions.

The first backup of any client is considered a "full" backup as TSM is considered a forever incremental backup solution. During the first backup the entire OS is scanned and sent to the TSM server storage. During the next backup TSM again scans the entire OS for changes and this time only backs up the files that have changed from the previous backup [incrementally on a day-to-day basis].



Archive

The archive process creates a copy of files and stores them for a specific time. Typically used for maintaining copies of vital records for legal or historical purposes. In the past backup and archive were interchangeable however within TSM they have become separate functions. Each object archived regardless of versioning will have its own entry in the TSM database causing bloat of the database and effecting performance. Typical running a backup with a longer retention period [via a management class] will achieve long term storage without bloating the TSM server database.


You can have different retentions for different data sets and can achieve the terms required by binding those data sets/directories, etc to a specific management class with the proper versions and retention periods.

 

[mgarczynski]

Thanks for the explanation. That does help. So if I want a point in time snapshot of my data, I should use an archive. That will copy the data and retain it until the expiration dates

 

[dms19]

Personally, I avoid archives. I'll run a different backup with the dataset bound to a longer management class to add longer retentions. If there are a lot of files, the archive will create an entry in the DB for each file under that client; creating bloat of the database. Multiply that by each client you want to archive and your database will grow that much faster over night. If the dataset is small, then you could use an archive but again its your preference, five different people will give you five different answers.