Summary |
Admin |
Home Page |
Forums |
Tracker |
Bugs |
Support |
Patches |
Wishlist |
Lists |
Tasks |
Docs |
News |
CVS |
Files
fbm - file and backup
management system
Origins of fbm:
The current version of fbm was developed for our own SOHO environment.
Full backup on CDs yielded stacks of coasters with multitude copies
of the same archive files, while need to limit this process resulted in insufficient
coverage of most recent or changed files. With HDD capacity of our system
going into hundreds of gigabytes we realised that something must be done
to provide fast, simple and well-controlled system. Instead of installation
of clumsy heavy-weight (and potentially unreliable) tape backup systems we
decided to create smart database system which will automatically assure that
there is always sufficient number of instances of each file available regardless
of storage media. Each instance is allocated reliability evaluation and
user can request different total storage reliability for various data.
At the same time we realise that many small and even medium sized companies
are struggling with the same problem. Stacks of tapes of unknown reliability
are shuffled, directories on servers are changed again and again to match
them to capacity of most recent backup device and boxes of CD coasters grow
in the cabinet. Resulting backup system is costly but seldom as reliable as
management and sysadmins hope.
We feel that we are not only ones who need something better.
fbm hopes to provide modular software system
for backup of any type of computer stored digital data. Comparison of fbm and traditional backup
shows main differences.
Basic design:
Basis of this system is database of all instances of every file. fbm
evaluates if every file is stored sufficient number of times on various media
and creates copying requests on as needed basis if insufficient coverage is
detected. fbm will also recreate damaged sections of working storage
media if required using information in its database. fbm can be used
to locate spare copies of individual files, or recreate
older versions of file clusters. Main differences between between fbm
and generic backup are listed in
fbm is based on PostgreSQL and the current version of scripts is
written in python. The current early version is tested (and used) on Debian
Woody Linux but working principle is platform independent.
System as planned will allow user to request specific relability and will
achieve it by combining old and new copies of files, deposited on various
media. Database of file instances becomes control center which can use old
copies on CDs, new copies on DVDs, tape backups, spare space on various networked
HDDs , USB based storage etc. There is no practical limit to total capacity
of data and almost no restrictions as far as type of storage media is concerned.
Old backups created by previous systems can be used if file format can be
read into database, including ad-hoc created stacks of CDs or even floppy
disks.
There are no real technical problems with implementation of fbm. Basic scripts
already exist and can be used for trials (creating very useful backup in the
process). Many more scripts will be needed to care for other types of hardware
and for other more advanced tasks. GUI control center may complement command
line interface at a later stage.
Current situation:
A few scripts exist which are currently used to keep our backup
system under control. There is little or no sophistication but even at this
level advantages are obvious and safety of our data significantly improved.
Full system will require many more tasks - see task
list.
$Date: 2003/12/20 06:43:52 $