Bergen's CTO: Why we moved to Linux

The second largest city in Norway is moving its servers to Linux. We chat to Bergen's CTO on the practical business of going open source in public

The City of Bergen recently announced plans to migrate many of its servers to SUSE Linux. We spoke to the city's CTO, Ole-Bjorn Tuftedal, to get his take on why Norway's second-largest city was making the move.

What made you initiate the change?
Most of the major Norwegian cities have economic problems, so it's always important to us to see if we are delivering our services in the best possible way. We had several of our systems and server platforms reaching the end of their support life, and we needed to do something about them. We were looking at moving to 64-bit platforms, and moving from proprietary processor architectures to Intel, so after some reviews and lot of studies we saw that Linux was an interesting alternative, it's reached maturity and stability in comparison to many other systems. We are doing a server consolidation project. We've got Oracle running on HP-UX and then will go on to look at our network servers, mail servers, DCHP, many of those which are already running on Linux, but may be running on different versions - so we are consolidating all of them partly to new hardware and to SUSE Linux.

Can you give me an example of the kinds of applications that students and citizens will be accessing with the new system?
We can roughly divide our infrastructure into administration and education. The administration is the standard one for the city employees, so all databases for the system used by the city administration, including health care, are included in this. Depending on how we count, there are between 20 and 30 servers that we hope to migrate to approximately ten HP servers. For the educational networks, we have 32,000 students and pupils, first through tenth grades, and up until now we have 100 schools, and each has had its own NT server. We are moving this, which is totally decentralised, to be centrally managed from our computer centre running on IBM blade servers. We are moving across server applications for students, which will be things like email, home directory, filing and print services, and Web services. The network is standardised on NT 4.0 with Windows 98 clients; we are moving to Linux servers and Windows 2000 clients.

Presumably the replacement of 100 Window applications servers with 20 Blade servers is the area where you'll reap the most benefit in terms of cost and convenience?
I am not sure yet where the main cost savings will be: we have done a couple of pilot studies with a couple of pilot schools, but we haven't done the main project so we still don't know all the costs. There are several different areas of cost reduction: we reduce the number of servers, which in itself is quite major, we centralise the management from 100 locations down to two boxes in a data centre, and of course there are reduced costs for server OS licences. We have achieved reduced cost for our support contracts, we expect to have a thorough evaluation in the autumn, when we will compute all the different aspects. On paper the savings are considerable, but we don't want to give away numbers.

What sort of ROI over what period are you looking at?
Due to the great reduction in complexity, I would expect the server investment to have quite a short-term payoff, but we are also replacing 2,000 to 3,000 PCs, which is a major cost in all of this. But then teachers and students get much better services. For example, each class has had only one shared email address in common -- but now each student will get an individual address that they can keep for the 10 years that they're in school.

Were you concerned about service and support?
It's important to get it, but we weren't in any doubt that we could get support, as we have already made agreements through several of the companies delivering us servers, and there are also several companies specialising in servers in general and Linux in particular.

Was your existing infrastructure giving you reliability problems?
The educational network had lots of troubles -- our database servers weren't unreliable, but they were quite old, the HP 3000 running older version of the HP UX, and we had several servers that had reached the end of life. After five years, hardware support costs rise. It's important you have full support for databases.

Has anything about the process surprised you?
We haven't had any unpleasant surprises, other than the press coverage. Despite some press reports in the Norwegian papers, we are not chucking out Windows. For our employees used to working on Windows and Office, to open the paper and see incorrect reports saying that we were terminating use of Windows caused quite a bit of discussion. We have already paid for Windows, so it would not be a good economic decision to jump about. Sooner or later, our Office platform will reach the end of its life, and we will re-evaluate what we do. We will look at a Linux desktop next year, but we have to test it, and it would mean that the teachers would have to use both the pupil's Linux desktop and their own Windows desktops, which might make life hard for them. Sooner or later we will review that, but there are no decisions there as yet.