Inline with my previous post “The simple value of archiving…” I thought I would post a CAS (Content Addressable Storage) overview because again it was something that I presented during the “Evolution of Disaster Recovery” seminar series that I felt we could have gone much deeper. Unlike traditional storage “Content Addressable Storage” (CAS) uses a content address or globally unique ID to represent the binary contents of the file, location of the file, etch… in contrast traditional file systems such as UFS and NTFS which use file names to identify files. A content address is typically calculated using an algorithm called a file digest (i.e. MD5, SHA-1, SHA-256, SHA-512) and a hash is created to identify the file contents.

Exercise 1:
If you are curious to see how hashing works you can download md5sum from here , Next open MS Word, Notepad, VI, etc… and type something and save the doc as “filename.doc” now open a command prompt (DOS window) and run “md5sum filename.doc” this will return something like this “b3a6616fb5cee0f1669b1d13dd4c98cb *filename.doc” now open the file “filename.doc” and change a couple of letters do not delete or add characters because this will change the file size (makes the demonstration a little less powerful). For instance if you typed “Hello dad” change it to “Hello mom” and save as “filename.doc” the file should be identical in size to the previous version, now run the “md5sum “filename.doc” the output is a globally unique identifier and it is different because it is examining the binary makeup of the file not the file name, location, etc… Right now on you file system you only have one document called “filename.doc” which contains “Hello mom” the version containing “Hello dad” is gone. If this had been saved to a content addressable storage device both instance would have been saved because although they have the same name they are in fact unique. After this excise you can probably see the value for compliance, corporate governance, revision control, etc…

Exercise 2:
Create another doc named “filename.doc”, run “md5sum filename.doc”. Now copy “filename.doc” to “filename.doc” and “filename3.doc”, run “md5sum ….” on these two new instances of filename.doc you will notice that the hash is identical. On a traditional file system we have consumed 3x the space required because the only identifier is the file name which is unique, on a CAS device the file names would be stored with pointers to “filename.doc” this is what we call single instance storage. The practical application of this scenario dramatically reduces storage capacity required by dramatically reducing the amount of duplication present on most traditional file systems.

The following graphic is a simplistic representation of how “Content Addressable Storage” works:

Hope this adds some additional clarification to the discussions we had at the seminars. If you have an comments, concerns, corrections or questions please comment on this post.

-RJB

I am flying from the east coast to the west coast to close out the “Evolution of Disaster Recovery” seminar series, with show 7,8 and 9 in LA, Orange County and San Diego respectively. As part of the seminar I spent a significant amount of time discussing Backup, Recovery and Archiving (BURA). I thought it might be a worthwhile endeavor to document this a little further and seeing that I have 6 hours to kill there is no time like the present.

The following example assumes a 5 week backup rotation, which is fairly common. In the event that you have a backup rotation that archives monthly backups for one year and yearly backups for 7 years, etc… the model grows exponentially. Typically most backup policies consist of incremental backups which are taken Monday through Thursday with a weekly full backup on Fridays. Tape sets are vaulted for 4 weeks, on week six week fives tapes are recycled, this rotation maintains 4 offsite copies at any point in time. Assuming a typical weekly rate of change of 10% data duplication is massive which extends backup times and raises cost due to the amount of media required. The following is a graphical representation of a typical 5 week rotation:

By introducing an archiving strategy we can greatly reduce the amount of data that is duplicated in the offsite facility and remove stale or unwanted data from the production data set which greatly improves backup and recovery times. The archive is an active archive which means that archived data is replaced by a stub (not a shortcut, stubs are not traversed during backups) and moved to an archiving platform of choice such as ATA(Advanced Technology Attachment), NAS (Network Attach Storage), CAS (Content Addressable Storage) , tape, optical, etc… - The user experience is seamless. A sample of what an archiving strategy might look like is represented by the following graphic:

Some duplication will continue to exist due to the fact that we may have frequently accessed data that we choose not to archive. The archive is static, any data that is read or modified is pulled back into the production data set thus there is no need to backup the archive on a daily of weekly basis. We refresh the archive backup following an archiving process which in this example takes place monthly.

-RJB