Three Windows to Linux Migrations (and Vice Versa)

daria42 writes "In this extended article, ZDNet Australia goes under the hood of three enterprises that moved their back end servers from Windows to Linux and open source software. Two of the companies ended up eventually going back to Microsoft, with the third one still going strong with Linux."

Comments

There and back again... A Hobbits tale

[CastrTroy]

This sounds like a pretty expensive procedure of going to Linux, and then eventually switching back when you find out that it isn't working for you. I think that with the right people, with the right knowledge, that it would be a good change for those involved. Buy you don't have to move everything all at once. Maybe just stop using windows for new things, and then eventually move the old stuff, or not. There's no reason to take down a working server, and try to replace it with something unfamiliar. replace little things, one at a time, and keep what's working for you. If you try Linux mail servers, and you just had a better time with Exchange, then leave exchange working. But if your database servers are performing better with Postgres, then leave that in place. There's no reason why you can't have a mixed environment.

Re:There and back again... A Hobbits tale

[bersl2]

I don't think that three is a sufficiently large sample to reach many conclusions in general.

Re:There and back again... A Hobbits tale

[hattmoward]

In part, I disagree. In my opinion, the first two companies were examples of what can happen when you try to build *any* environment when your techs don't have enough knowledge to do it right. Windows can have terrible identity management too, if someone doesn't get it set up correctly. Exchange servers can be integrity and uptime nightmares in the wrong hands. Right now, though, you are definitely going to find business applications that aren't available on Linux, and if your techs can't find a way to make it work or find something similar, you are SOL. If it's something you can't do without, well Windows may be the only solution.

What works best

[Billosaur]

So we have a breakdown of 3 companies, 2 switching from Linux to Windows, 1 from Windows to Linux. Is there any great wisdom to be gleaned from this? The only bit I can come up with is that you use what works best for you with the infrastructure you need to support. It's easy to say Linux will work well for everything but that's just not realistic. It's also safe to say that Microsoft sucks universally, yet there are plenty of sites running SQL Server and IIS that seem to be doing ok.

If you're smart, you analyze your needs and then add 50% for growth and ask yourself if the infrastructure and technology you plan to use can handle it. It's simpler than getting caught up in the Microsoft vs. Liunx battle for supremacy.

Re:What works best

[hey!]

Is there any great wisdom to be gleaned from this? The only bit I can come up with is that you use what works best for you with the infrastructure you need to support.

I think it's interesting to contrast the foci of the switchers.

The Linux to Windows switchers were looking for an OS that was an all-around general purpose operating system to support a wide array of day to day internal IT uses. The problem they have is scaling up their Linux support to keep all the endless bits of software they organization needs working together. The "network effect" applies here: there's lots of companies with their needs and (low level of) resources, and those companies by in large use Windows. Therefore vendors in the Windows space address their needs to do a wide variety of things good enough with limited staff expertise.

The Windows to Linux switcher was looking for the best platform to host a single application they were designing. The consumers of the platform were, in effect, the development team, which was small and a higly focused center of expertise. They are looking for maximum performance and stability to support a universe of software they define. Joe Blow in accounting having to put three passwords in to use the VPN is quite low on their priorities, compared to, for example, hitting an unanticipated wall in the performance curve.

In short, the Linux->Windows switchers viewed software as a support function -- back office stuff. The Windows->Linux switcher viewed software as a line function -- outward facing stuff.

Of course a sample of three is nothing. But anecdotally, it's intriguing.

Skill problems

[__aaitqo8496]

TFA exposes the "problem" in both instances that the company reverted back to Microsoft:

Lacking skill set

Under that scenario, any OS switch would fail. You can't blame Linux (or Windows or any other OS) for that problem. Linux should be viewed as Linux, not Another Windows. You need proper IT support.

Some of the problems were simply lack of knowledge. There were complaints of having to claim 3 passwords for VPN access and not utilizing a worldwide-accessible central information store. That's just laziness.

Re:Skill problems

[Reverend528]

I think this quote from the article really supports that theory:

"If something breaks in Linux and you've got the knowledge, you can generally fix it and get it up fairly quickly by yourself."

"The problem was that just one or two people in the group [out of 15 IT staff] could do that."

Re:Skill problems

[digitalgiblet]

"Their world is all windows and if you try to get them to open up to new ideas they put their hands over their ears and shout "Na, na, na, I'm not listening!!!"."

If you replace Windows and Linux in the previous sentence with any other competing ideas (or swap them), you will pretty much still have a valid point.

People who have time, energy and money invested in Windows aren't going to want to switch to Linux. People committed to Linux aren't going to want to switch to Windows. Or Mac. Or Amiga for that matter...

You can make the same arguments for religions, metric system vs. English, emacs vs. vi, republican vs. democrat, tastes great vs. less filling, etc, etc, etc.

Most people like the feeling of belonging to a group. Not all, of course, but most. There are many kinds of groups including the "outsiders" who like to hang out together because they don't feel they belong in other groups.

If I identify myself as a member of group A, it is very difficult to truly have an open mind about the beliefs of group B. Sometimes I even begin to believe that the members of group B are intentionally bad. Part of identifying what makes group A distinct is by contrast with group B. Our biology is wired to make binary distinctions about our environment (I eat it vs. it eats me). That makes for quick decisions. Our SOCIAL systems are almost always made up of a VAST number of shades of gray. A truly open minded person realizes that there are good things and bad things about both group A and B, and then wieghs which group meets their needs best, but acknowleges that for some people group B is a better fit.

For some people Linux is better, for others Windows is better. It is all in the definition of "better". If you want a stable, low-cost, open source system, then Linux is "better". If you want a system that runs specific software you need for your buisiness and it only runs on Windows, then Windows is "better".

It is much easier to judge an operating sytem based on your own criteria and then decide that the one that matches those criteria is empirically better. If you are a "technical" person whose criteria are solely technical then you are likely to arrive at a different conclusion from a "business" person whose criteria are solely business based. Neither of you is emperically right or wrong. Balancing and blending the two is the art.

Using Linux correctly?

[Rekolitus]

"Previously, all the e-mails were effectively stored on the desktops and there was no central location of the data,"

Perhaps it's just me, but it seems like these companies weren't applying Linux quite the intended way. Linux isn't a magic plaster you can throw over your IT problems, and frankly, I'm sure it does need a little more maintenance than Windows, but it seems like the people that set these systems up didn't put any thought into their infrastructure.

Storing emails on the desktop isn't a problem that Linux creates. Windows seems more akin to something that says "This is the best way to store emails", whereas Linux is more like "Where do you want to store emails? It's up to you. I can't give you any advice." I'm sure these company's Linux-based experiences would have been much better if they did a bit more planning into the structure of the services in the first place.

Re:Using Linux correctly?

[Chandon Seldon]

Linux should need less maintnence, but it may well need more setup time.

David Braue

[dajobi]

Anyone know anything about the author of this story? That first story, the one about Austereo, looked kinda schilly to me, especially comparing what the author wrote to the quotes from the interviewee. Compare:

After three years with Linux, however, Austereo began reconsidering its choice as continued growth in the company led to increasing complexity within its IT strategy -- exposing the limitations of the Linux-based environment in some very painful ways.

This time around, things are working much better, with a full range of Microsoft server applications providing a deeply integrated, highly effective IT infrastructure that has significantly improved productivity and transformed information management within the company.

Working with Microsoft consultants, he and his team sat down to map out their future infrastructure and found that their requirements could be easily met using an integrated suite of applications built on top of Windows Server 2003.

After thorough discussions, Austereo committed to a move away from Linux onto an architecture combining Microsoft SharePoint Server, Exchange Server and SQL Server as well as Office 2003 and BlackBerry-related add-ons like the BlackBerry Enterprise Server. It was a hard decision, but even Forgione concedes he was impressed when comparing the company's existing and potential computing environments.

To:

"Importing our network environment and applications onto a new platform required some fairly specific skills, and those skills were not abundant within the group. As the business started to grow and we realised we needed to provide additional services to help people accomplish their day-to-day tasks, it became a very obvious and glaring issue."

"The problem was that just one or two people in the group [out of 15 IT staff] could do that, and it was hard finding people who understood that [open source] isn't just about playing with these tools, but delivering something."

Braue's version: Linux wasn't good enough, a Microsoft "solution" was required.

Forgione's version: The IT staff didn't know how to use Linux. For some reason we didn't think hiring competent staff would be a good idea.

What do you think?

Re:David Braue

[RogerWilco]

Working with Microsoft consultants

That's a significant part of the 'solution', they seem to say that they didn't have the knowledge inhouse to design an infrastructure that met their increasing demands, and they had 'MS consultants' to turn to for a solution.
I think there might just be to few 'Linux Consultants' that are capable of providing this kind of service, even if the tools might be out there. Pitching them to management in the right way is a skill in itself.

Re:David Braue

[swv3752]

The way I read it is that the Linux people they found would not rush something out the door. You typical Linux admin will want to go thorugh a rigourous testing phase and write up some scripts to automate all the regular process, while a Windows admin will slap something together and shout "It boots so it works". Managers are often pennywise and pound foolish so wil go after the Windows option.

Additionally, they probably had unrealistic expectations of expertise. Most Linux admins would be willing to learn on the job, but you need to allow a bit of time to learn. Of course, thier current admins could have RTFM to learn on the job, but...

Stockholm syndrome

[metamatic]

There's nothing new about the phenomenon, either. I remember research back in the 90s showing that Macintosh users were all familiar with Windows, but that Windows users were mostly completely ignorant of the Macintosh.

In other words: Mac users who said that Windows sucked, generally did so from a position of knowledge, whereas Windows users who said that Macs sucked, generally did so from a position of ignorance. I expect it's still the case today, and I wouldn't be surprised to find out that statistically, it's also the case with Windows vs Linux in the enterprise.

My conclusion: it's all about services

[brokeninside]

In the linked article, 2 out of the three companies migrated to Microsoft solutions based on services available from Microsoft because their internal IT departments didn't have the requisite knowledge to keep entirely open source solutions up to speed with regards to internal growth. The third company remained with Linux as a result of Linux services offered by Sun as part of a package with new Opteron based servers.

The other interesting bit is that a key part of decisions made in all three cases was the available software. The first two companies went with Microsoft because of Sharepoint. The last one stayed with Linux on Sun hardware because of 64bit J2EE.

Re:Why does it have to be either/or?

[denverradiosucks]

I submitted before finishing my comments. Any sort of penetration into the Windows world through Linux is a great step in the right direction. It's just not ready to replace windows on all levels yet.

A revealing number

[romrunning]

In looking at the numbers, both of the organizations that went back to Windows had over 1,000 employees, and the other example listed had a little over 100. Both "back-to-Windows" orgs cited complexity as one of the reasons for the return to Windows. Is it possible that Linux developers have been too focused on each individual app rather than how the apps interact with each other? Who is responsible for the overall vision?

Re:A revealing number

[51mon]

I don't think so -- I think it is simply dearth of experience in the area.

Whilst ISPs have been using free software solutions for user management for years, and love it because they can easily integrate any old third party software without coughing up money. You could probably group all the people who deployed such in a small stadium.

Skill in big directory services type skill on Unix/Linux is pretty sparse on the ground. Probably a lot more people like me around who've done integration with relational database for authentication and authorisation, done NIS, done NIS+, but look at LDAP and Kerbero,s and go "wibble wibble argh". I do PAM now and then as a Linux admin, but never frequently enough to say I'm know it, it is set and forget for most small companies, and most specific server set ups.

Heck even integrating "samba" into ADS has a pretty big learning curve. On the upside the people who have done that understand what the Microsoft systems are doing underneath (often this is the reason they are die hard GNU/Linux bigots).

I think this is reflected in the story, the company which is doing the big service for lots of users, loves the "we can roll our own solution" approach, as they only have to do it once for 2,000,000 users, for a limited range of apps.

The others are trying to find the skill to do these kind of things. Although if someone stuck email on the desktops in 1999 they were pretty lame, I was deploying IMAP4 on my home network by then, to avoid keeping email in proprietary formats on the desktop.

On the other side of the coin, whilst Microsoft have put some nice GUIs on top of ADS, the number of people who really understand Microsoft ADS and the security model is surprisingly small, and clearly didn't include a lot of people who worked on the built-in services for Windows 2000 at Microsoft. So a case of "its fine when it is working".

I'm very surprised at the comment about having to spend time managing GNU/Linux systems. Our experience is that it is the Windows boxes that "suddenly" do something odd, because of the lack of transparency, and poor logging (what you mean it can log filesystem corruption without mentioning which filesystem got corrupted ?). Similarly surprised at comments on server consolidation, my experience has always been that Windows almost invariably ends up being deployed on more boxes, either for performance reasons, or to separate services because of the frequent needs for reboots when applications are updated.

I suspect a lot of the "benefits" here would have accrued, be throwing out the old, and bringing in the new, whatever the new was.

If you...

[jd]

...change "setup time" to "setup thought", I would completely agree. The problem with Linux is not the time to set it up (which is roughly comparable to Windows, sometimes less), but rather the time it takes to figure out what you actually want.

A simple example would be deciding on your e-mail system. Sounds easy, right? And it is. If you know - in advance - what sort of e-mail system it is you actually want. Just saying "e-mail" doesn't tell you very much. If you need a great deal of power in the mail processing engine, you're probably going to want Sendmail. If you need to blast through vast quantities of e-mail very quickly, Postfix is a better bet. If your company is relying on Exchange services, then you're looking at something like Open Groupware. If you aren't using Exchange clients, but do need similar services, then OpenXchange might do what you want.

That's just for e-mail! Then you have to think about all other intranet services, which have a similar level of flexibility. Internal web services with static web pages will be better off driven by Tux. Java servlets, these days, really mean Apache, as they're the ones mostly working on that capability. Basic scripting with reasonable power and reasonably dynamic content would probably mean Roxen.

If you want virtualization, you've three entire tiers - total machine simulation (vmware), heavyweight containers (xen) and lightweight encapsulation (vservers). If you want to admin the box, do you edit the config files, use Red Hat's scripts, use Linuxconf, or use webmin? And the list of options goes on and on and on.

On the one hand, the choices give an aware user a fantastic level of power and almost superhuman control over their system. On the other hand, it means that you cannot approach this with a turnkey attitude. This should be no great surprise. You can drive a roadcar with a turnkey attitude and expect to get from A to B in one piece. This isn't going to work in a Formula 1 racing car or an X-15 experimental aircraft. Why should it? If you act as though these are all one and the same, your efforts to transfer over WILL fail. This is not a limitation of these vehicles, it is a failure to recognize that simplifications that are true in one case won't hold for the general case.

Let's look at one of the big complaints I've heard for Linux - a lack of wireless card drivers. How many of those who are complaining have actually looked for additional drivers? My guess is that half the complainers have not, and that the majority of those would find that a project just as madwifi would provide the drivers they want. There are a few others listed on the Linux WPA Supplicant page. "But we don't want to install 3rd party drivers!" That wasn't the complaint - the complaint was that the drivers didn't exist. If I can find the drivers, and they DO exist, I will have zero sympathy for those who then come up with further excuses - because if the complaint has to change each time it's proven wrong, then all it is IS an excuse.

My guess is that almost every single case of a company "needing" to switch from Linux to Windows will - on closer examination - prove to be a case of nobody bothering to figure out what the company actually wanted, OR nobody bothering to figure out how to get Linux to provide it. There will be VERY few cases - although such cases will happen - where Linux really isn't a good fit, which is a limitation of Linux, but I seriously doubt that more than one in a thousand migrations from Linux to Windows fits into that category.

Bad planning and lack of skilled people

[the_arrow]

After reading the article (yep, I did read it), it seems to me that for the two companies that switched back to Windows, the Linux-switch was not very well planned and the needed skillset was not there. The switch back to Windows was on the other hand very well planned.

You don't know what you don't know

[WebCowboy]

Their problems were that they wanted features they weren't finding in Linux, but did find in Windows.

WRONG. The problem was that they weren't aware that Linux-based systems acutally COULD do what they wanted and a Microsoft marketroid came along and showed them how it could be dome with Windows. Both examples of the reversion to Microsoft showed all the hallmarks of "you don't knw what you don't know"--not only did their IT people not know how to make Linux work for them--they weren't even fully aware of the capabilities of open source systems.

The articles mention one comany migrated to Linux 3 years ago, the other seven years ago. Did they really keep up with the fast-moving world of open source? In the anecdote about the company that stuck with Linux there was a fairly siginificant mention of upgrading both hardware and software, but in the other cases little to nothing was said about upgrading. It is entirely possible that the latter company was still runnning on their original Mandrake 6.x (or whatever it was in 1999) platform. Were they expecting their Linux systems to be magically immortal? If they implemented a Microsoft system in 1999, do you really think they'd be happy with NT4 and the big pile of manure that passed for Exchange Server at the time? I seriously think not.

I think the final solution of migrating (back) to another platform was too drastic, and that these companies dropped the ball when it came to examining the open source alternative. Three passwords to log into a VPN? Email boxes stored on clients? Lack of collaberation tools? COME ON! You can set up a Linux server to allow a Windows client to log in without any extra passwords. It isn't hard to set up a secure IMAP server using Postfix to manage mail server-side either, and there are "Exchange replacements" that may fit the bill if you need to do mare than just manage email centrally. There are a bazillion "portal frameworks" out there, and Subversion can be used as a collaberation tool for more than just computer code. I know this can all be done because I've done all of that myself. These people are lazy and uncreative and didn't even try to find a more elegant approach to solving their problems. Instead they let a Microsoft salesman sell them a sledgehammer to drive in their 10-penny nails.

These stories also underscore a problem with the Linux community as well, however. Microsoft made themselves readily available. They have an education programme that turns out MCSEs faster than rats can breed. The Windows brand is everywhere and they make it very clear with every release "what's new". Where were the Red Hat and Novell people when these Linux shops were struggling? Why isn't red-hat more agessively marketing and expanding RCHE certification? What about LPI? And as far as marketing goes, IBM has done a bit but Linux is far from front and centre, and the marketing presence of Red Hat and Novell is next to nothing in comparison to Microsoft's mega-campaigns that contain heavy dollops of information (or mis-information in some cases). Yes, MS is the big man on campus and has the resources to pull all this off the best, but it's going to take a huge marketing and support effort by the Linux community to make sure we not only convert more people to Linux but to retain them as well.