Interview: Kern Sibbald

Kern Sibbald is the project founder and lead developer on the OpenSource Bacula network backup project.

He holds a PhD in Solid State Physics, but has always worked in the IT industry. In the early eighties he co-founded Autodesk (the maker of AutoCAD). Since 1999 he has devoted full time to Open Source projects, first with apcupsd (UPS management program), then with Bacula (network backup).

His talk will present Bacula and the project around it.

What's the aim of your talk?

Mostly, I hope to make more people aware of what Bacula is and what it can do.
Secondary considerations are to attract new Bacula users and Bacula developers.
Finally, since I have been working full time on Free/Open Source projects for the last 8 years, I would like to meet a few of the Bacula users as well as other Open Source programmers.

Have you attended previous FOSDEM editions?

No, this is my first Free/Open Source conference.

Bacula uses a modified GPL v2 license. Which specific problems did the original GPL license expose?

The GPL did not really pose any specific problems. However, I have added four clauses that clarify certain points:
1. to permit linking Bacula with Open Source programs such as OpenSSL that are not considered to be GPL compatible.
2. A clarification of IP rights (i.e. despite our best effort we cannot guarantee that we do not violate any IP rights).
3. Each contributor to Bacula certifies that he has the right to grant the license.
4. For the Windows version, we use some Microsoft source code that is publicly available, but which we do not have the right to re-distribute except in our binaries (i.e. the user must get the source code for those particular parts him/herself).

How were the reactions of the community after the 2.0 release?

They have reacted very positively. Judging by the support requests, we have picked up a good number of new users, and I was also surprised at how many current users immediately upgraded.

Bacula has become very popular in a short amount of time. Can you sketch the progress since the early days ?

Oh, to me, Bacula has been around a long time -- 7 years now, and it has been released to the public since April 2002. However, what you say is true. Until recently, Bacula was adopted slowly.

Here are a few key dates:

  • January 2000 – Project started
  • 14 April 2002 – First release to SourceForge (version 1.16)
  • 29 June 2006 – Release 1.38.11
  • January 2007 – Release 2.0.0
and here are a few key usage statistics:
DownloadsVersionYear
741.162002
10,9191.362005
12,4081.38.112006

Although these are not large figures compared to many Open Source projects, they seem to me rather substantial for a backup program.
According to SourceForge statistics:

ProgramDate RegisteredDownloads
Amanda1999184,681
Bacula2002338,368

I don't know exactly how they measure these statistics (I assume it is the total number of downloads), but I am impressed and pleased with Bacula's popularity compared to such a well know backup program as Amanda.
There are other Open Source backup programs (BackupPC, ...) which have much higher numbers but I don't think they compare in functionality to Bacula and Amanda.

Bacula's SourceForge page hits are increasing rather dramatically from year to year.

What kind of "enterprise" features does Bacula have today, compared to commercial offerings ?

Quite a few. A small list is:

  • Volume management
  • Pool management
  • Data migration (disk to tape, ...)
  • Data spooling
  • Encryption
  • Multiple simultaneous jobs
  • Data may span multiple volumes
  • Support for many OSes (Linux, Unix, OS X, Win32, ...)
  • Autochanger support
  • Support for most devices

See: http://www.bacula.org/dev-manual/Current_State_Bacula.html for more.

Speaking of advanced features... Is a "snapshot filesystem" (like e.g. ZFS has) possibly a useful feature for backup systems like bacula?

Yes, we use Microsoft's Volume Shadow Copy Service (VSS). On Unix/Linux systems snapshot filesystems are not in hot demand.

What's your personal backup strategy like ?

  • Backup everything on my server to tape every night (was DLT-8000, now LTO-2).
  • Backup the critical data on all my other machines nightly.
  • Full backup once a month.
  • Differential (all changed files since full) once a week.
  • Incremental every night.

All backups go to the same tape that is retained at least one year.

For a site that I administer that uses disk based backups with 3 different pools, I use the following policy:
Maintain 1 year of Full backups, 6 months of Differential backups, and 4 weeks of Incremental backups.
This scheme is documented in one of the chapters of the manual.

Better be safe than sorry! Thanks for taking the time to answer our questions.


Additional links:

Creative Commons License
This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.