Data Backups

Why backups are performed

Server backups are carried out mainly for the purpose of disaster recovery. Users are therefore responsible for their own data, the accident can not be an excuse for the students to not submit the project, and the data need not be up to date in case of recovery.

However, if someone loses something in any way (e. g. a mistake delete or overwrite something), you may contact the Rudolf Čejka, preferably using an e-mail, if it is possible to restore the data. But it is always necessary to precisely specify, what is to be restored and from when.

When backups are performed

Each backup starts at 1:05 and the day determines the type of backup: In 1st Saturday of the month will start full backups, in other Saturdays will start differential backups against last full backup, and from Monday to Friday every day will start incremental backups against any last backup.

Hidden hazards of differential and incremental backups: For backups, that are not full, the files are stored only if they have newer modification time than last backup. So if someone creates a file with older modification time (e. g. typically extracting archive files *.zip, *.tar, etc.), it is not included in these backups. Another problem may cause the deleted files, which are not identified in differential nor incremental backups. During data restoration for a particular day all files from the last full backup are created and then newer files are updated or added.

What is backuped

All important servers are backuped. From the user aspects, the most important are kazi.fit.vutbr.cz (employee home directories and e-mail), eva.fit.vutbr.cz (student home directories for the FIT and e-mail), aja.fit.vutbr.cz (employee home directories in Windows), fik.fit.vutbr.cz (student home directories for the FIT in Windows), fest.stud.feec.vutbr.cz (student home directories for the FEEC and e-mail), tereza.fit.vutbr.cz (web server), agata.fit.vutbr.cz (database server) and merlin.fit.vutbr.cz (student server).

As of 1. October 2010, there are regularly backuped 30 FIT servers, 5 FEEC servers, and 1 remote server located in Prague. The total amount of data exceeds 5 TB per one full backup.

How backups are performed

To backup is deployed open-source program Bacula (2.4.3) with GPL licence.

Where backups are stored

Backups are stored on one of 2 LTO-3 tape drives with 400 GB capacity without compression per one tape media. In actual operation, there is often stored about 600 GB per one tape media by hardware compression contained in the tape drives.

Tape drives are located in Overland Neo 4200 tape library with 120 tape slots, automated robot, and barcode reader for quick identification of individual tape media.

Backup speed is highly dependent on a particular server. Limiting factor is not usually connection speed to the backup server (100 MB/s) nor writing data to tape (40-100 MB/s), but rather the speed of reading data from disks (1-30 MB/s).

Back to Important Information and Guides