Home > Ubuntu, Ubuntu Server > What I want Ubuntu Server to be

What I want Ubuntu Server to be

What we call “Ubuntu Server” is currently a basic system, released together with a giant set of packages that you can install. Being Debian-based, the result is very close to what Debian also delivers: a powerful base platform that can be customized to do almost anything. We add some killer features and packages that we feel are missing, do some integration and bugfix work, and then release in a predictable schedule.

But I want Ubuntu Server to be different. Ubuntu Desktop was successful by releasing a product that delivers a cool, integrated, user-friendly desktop experience. Ubuntu Server should be a product that delivers a cool, integrated, sysadmin-friendly server experience. I want us to innovate. I want us to boldly go where nobody has gone before. I want Ubuntu Server to be something, rather than be anything. We should have a product that implements industry best practices, rather than just making them possible. Having configuration files under a revision control system, users always defined in a local or distant LDAP directory, screen on network shells, LVM always configured… We should deliver and integrate the right software solutions rather than just making all options available.

Yes, that means making some tough choices. That means having a bigger out-of-the-box install. We must recognize there are two separate use cases, resulting in two separate products. You might want a minimal system and install/pick only the stuff you need. Or you might want something that helps you getting things right. A Product rather than a Platform.

But wait… That’s already the case: we have a minimal/appliance installs, separate from the “server” install ! So all what’s left is to operate a semantic shift in our current users: what you called Ubuntu Server so far is really an Ubuntu minimal platform. And we would continue to deliver that. But Ubuntu Server could be something different, something where we make more choices on behalf of the user, in order to deliver a better and more integrated sysadmin experience.

And let’s get ready to take some heavy flak for the choices we’ll make ?

About these ads
Categories: Ubuntu, Ubuntu Server
  1. T0m
    March 20, 2009 at 10:33 | #1

    Differentiate or die!

    I 100% support your idea. ATM Ubuntu server is too similar to debian. A longtime debian user won’t change to Ubuntu. Same for a Red Hat/Centos user.

    But with your ideas that could change. I think it is really worth to try and it would be very in line with the Ubuntu desktop. You have to pick favorites.

    If you pick the right ones the majority of people will like you.

  2. Roger
    March 20, 2009 at 11:06 | #2

    I also agree 100% – this is exactly what seems to be missing from the Linux space at the moment.

    Running Linux servers requires a much higher level of knowledge to start with than doing the equivalent on Windows. This is fine – Linux is a better system IMHO – but it isn’t very accessible.

    A small IT shop can buy Windows SBS install it and expect to set up Active Directory, DNS, DHCP, Mail, File Serving, Print Serving, etc.

    There are products aiming at this space in Linux, e.g. Clarkconnect, but they are generally offering a fixed set of functionality and are not designed to be as flexible as Debian or Ubuntu or another general purpose distro.

    Ubuntu has done a fantastic job with the desktop, creating something anyone can install and use and, importantly, grow with.

    I would love to see something similar in the server space. A system that gives you simple choices to make, and them installs a system that meets your needs and implements current best practices, so that once you have learnt a bit more you don’t have to go throw it away and start again to get LVM, LDAP, etc.

  3. aim
    March 20, 2009 at 11:21 | #3

    The idea IS cool. BUT – if we have such a product this could lead to the situation where we could found ourselves between numerous “sysadmins” who can only say “make everything cool” and does not know all there is to know of this “coolness”.

    My opinion that the server team should concentrate on server use cases and document each as accurate as we can. So the distribution, ubuntu server would be what it is now – a small base system. And all the work should be done using documentation.

    This is how we can hold a balance between “magician computing” and “science computing”.

  4. Stoffe
    March 20, 2009 at 11:43 | #4

    While I personally like to command-line manage any servers, I do think that there is something to be said for graphical alternatives such as when managing Windows servers. That could be one way to go and differentiate, by serving both preferences, especially if they can be used intermingled, that is the tools work well together even if using both.

  5. March 20, 2009 at 11:43 | #5

    ++
    Make it easy to have a preconfigured webmin/virtualmin system so sysadmins to have a ready to go system for websites similar to cpanel/plesk whatever

  6. leonel
    March 20, 2009 at 12:10 | #6

    I guess the base system and task selection we got does a great job
    but I guess this step you propose needs to focus on a graphical administration ( web interface / xserver interface) for all packages available
    1st question : what to do when the package has already its graphical admin interface like http://www.cherokee-project.org ??

    What about ebox ?? I haven’t tried

  7. March 20, 2009 at 12:21 | #7

    Wow you really captured the spirit of Ubuntu on the desktop. Please do it, make choices for those who do not have the experience or time to make their own. There’s a big need for such a server product.

  8. martinbrook
    March 20, 2009 at 12:47 | #8

    This is a really great idea. At the moment Ubuntu server doesn’t have its own identity – to me, it feels just like ubuntu minus gnome.

  9. Matthijs
    March 20, 2009 at 12:58 | #9

    Very good idea indeed. If this solution will still be scalable, then I can’t think of any reason why anyone would _not_ want this :)

  10. anon
    March 20, 2009 at 13:29 | #10

    > LVM always configured

    Ugh. Please no. Any sysadmin worth their salt would know that LVM doesn’t support write barriers. No thanks!

  11. Roger
    March 20, 2009 at 13:29 | #11

    Aim said:

    > So the distribution, ubuntu server would be what it is now
    > – a small base system. And all the work should be done using
    > documentation.

    I respectfully disagree.

    I think that a server install should be configured with LVM and version controlled /etc by default (with an option to not do this during the install). This could either be part of the install, or part of a firstrun type program.

    I don’t think that anyone is suggesting that a default Ubuntu Server install will be much bigger than it is now – i.e. it should not install a whole load of server-type packages by default.

    Just that when you do install a package you get a much better configuration that you do at the moment. The Dovecot-Postfix integration in Jaunty looks like being a good example of this.

    Similarly, wouldn’t it be nice to install Ubuntu and get asked whether you want to use a) Local LDAP b) Remote LDAP c) Likewise Open or d) /etc/passwd for the user authentication.

    I would love to be able to install a file-server package (or meta package) and get quotas, acls, Samba and NFS all set up for me in a way that works together.

  12. jimcooncat
    March 20, 2009 at 13:32 | #12

    I think having a minimal install with metapackages is the way to go. Admins have varying tastes. LAMP (with Mysql & PHP) or LAPP (Postgresql & Python)? You can start some real flamewars by picking default choices.

    I’d love to see some special combination packages. For instance, a drbd-based home directory setup with two servers. Minimal installs on both, standardized IP addressing, then install drbd-primary on one, and drbd-secondary on the other. Bam! Instant fault-tolerant home partition. (hey, it’s just a dream.)

  13. Roger
    March 20, 2009 at 13:35 | #13

    Tom said:

    >Differentiate or die!

    If I want to set up an Oracle database using a clustered file system I’m going to use Red Hat.

    On the other hand I think that there is a real need for a Linux distribution that lowers the barriers to entry for running Linux servers and I think that Ubuntu is well-positioned to fill that niche.

  14. tana
    March 20, 2009 at 14:25 | #14

    I find aim’s suggestion very interesting. Of course, ubuntu server shoud make the installation of a minimal server possible. The next step is to work on use cases. These use cases should be documented. We shoud have the description of the case. We should explain how to solve this problem and how to do that with ubuntu server. Ubuntu server must provide at least text-based configuration tools to configure what is explained in the documentation.

    Even if Ubuntu server team has to add many many scripts or whatever to make these changes possible, sysadmins should not need to learn all of these programs. They should use something like “ubuntu-admin” and that’s all.

    The use cases should include a scenario where we suppose that everything that may be remote is remote.

    We should have also in mind that the service we are configuring can be clustered or load balanced…

    I like very much the idea of putting config files under a revision control system.

    Generally, we should have in mind that the server is not isolated. This could help people choose ubuntu server since they can do easily what they have to do in everyday life. We should not suppose that remote servers are also ubuntu servers.

    It would be nice to be able to make some monitoring easy with a well configured snmpd for example and a reporting tool (perhaps web based…). The same thing for supervision and backup.

    We should have something for virtualization. It should be easy to configure and use virtualization.

    Everyday LVM operations should be very simple. Some sysadmins are frightened by the idea of hot-changing the size of a partition if they have to use lvextend, ext2online, etc…

  15. oliver
    March 20, 2009 at 14:44 | #15

    Sure, go ahead and differentiate – if we as users don’t like it, we still have Debian as solid fallback :-)

    Seriously, if you want to make a difference, then decide which people you want to target, and find out what troubles them about current server systems.

    Personally, I usually have troubles with upgrading a server, and with restoring a server backup to another machine. So maybe some software to replace the apt-based upgrade system might be an idea: instead of running apt-get dist-upgrade and hoping that the resulting system will be OK, better create a list of packages, save related config files, tell the user about “non-managed” software (in /usr/local/) and non-managed config files, and then do a clean reinstallation of a base system + all those packages, and restore and migrate config files. Bonus points for automatically creating a full test system in Virtualbox first. Also bonus points for making upgrades of single packages similarly easy (I just spent a whole evening upgrading Django 0.95 to 1.0 on a Debian server and migrating all my Django apps to 1.0 at once – adding some safety net for a step-by-step upgrade would certainly have helped).

  16. jef spaleta
    March 20, 2009 at 17:16 | #16

    I think this is the first articulated vision I have seen which makes a coherent attempt to state a roadmap for what Ubuntu server is. It’s certainly not appropriate for me to nitpick the details..thats between Canonical and its paying customers. But, its refreshing to see any clear vision articulated that really tries to demarcate where a Ubuntu server product can bring value over a Debian server install.

    The “be something” instead of “be anything” idea probably has some real legs to it…even if the particular something envisioned here ends up being the wrong something. I think other commercial entities have already figured that out, and are already doing exactly that by basing their offering on Ubuntu server.

    Just be careful that other downstreams like TurnKey don’t out innovate you Canonical’s own turf. TurnKey is building a business out of taking the flexible Ubuntu Server as it sits right now and turning it into “something”.. multiple somethings in fact. If you change what Ubuntu server is, you will be disrupting that downstream commercial entity. I wonder how they feel about that happening. It will be interesting to watch the transition.

    http://www.turnkeylinux.org/faq

  17. March 20, 2009 at 17:19 | #17

    leonel had an interesting point. Just before your post on Planet Ubuntu was this: http://ubuntu-tutorials.com/2009/03/19/configuring-printers-via-the-cups-web-interface/

    There was also this, earlier: http://pthree.org/2009/03/13/use-the-tools/ Aaron Toponce says, “Whenever you’re editing config files, and a tool exists to make the change, use the tool instead of editing the config by hand.” I agree, and observe that ‘the tool’ doesn’t necessarily have to be a command-line program, but could also be a GUI.

    But he also says: “If the tool doesn’t succeed, learn why.” I used to really like cPanel and WHM for administering shared hosting; now I hate them. They give you an illusion that everything Just Works, but when they fail, they fail in a very non-transparent way.

    Given Ubuntu’s history with the desktop, it could probably eventually include a server GUI that is transparent, simple, modular, degrades gracefully, etc. But, I have started using Dustin Kirkland’s screen-profiles package, which is wonderful, and a step in the direction you are referring to…and when you consider the work that goes into it (making screen “just work” for servers) it makes a web-based GUI seem like a *lot* of work.

    The commenters above talking about “use cases” should really look at tasksel. Aptitude lets you install or uninstall the tasksel tasks defined for Ubuntu. This could probably be extended to cover what you want.

  18. March 20, 2009 at 17:39 | #18

    LVM supports write barriers in 2.6.29.

  19. Michael “Cool” Howell
    March 20, 2009 at 22:20 | #19

    Truly, Ubuntu Server should have a goal ahead. Just some nitpicks of what I think should and should not happen.
    1) Ubuntu Server should not come with X. Use a web interface or a good menu-based CLI.
    2) Ubuntu Server should be easy to set up.
    3) Ubuntu Server should have an easy-to-setup CMS (for web servers).
    4) Ubuntu Server should not come stock with any particular server software. The more software you have, the easier it is to be compromised.

  20. kragil
    March 21, 2009 at 12:41 | #20

    @jef: Thanks for the link. I think all the appliances should be install options in a Ubuntu server install. Maybe the server team should work together with the turnkeylinux community.

    And as always I don’t like your negative tone. It is all about making the pie bigger and not about slicing it into smaller pieces. But that post was very good (for your standards :P )

  21. jef
    March 21, 2009 at 18:16 | #21

    kragil,

    The question is, how important is the Ubuntu server offering as a revenue stream for Canonical moving forward? Can Canonical work directly with nother small business entities like Turnkey who may in the future compete directly with Canonical via support services in the cloud?

    Ubuntu isn’t Debian, the directions the official Ubuntu server can go and the shapes it can take are controlled to a large degree by a single corporate for-profit entity, Canonical. Is Canonical willing to give up some control to another for-profit entity like Turnkey can take a lead interest in shaping the official Ubuntu server? Is Ubuntu a big enough tent for multiple software business entities to have a seat at the table to design and build the official Ubuntu official offerings?

    From a community perspective, it’s great that Ubuntu can be a platform for other businesses to innovate. But its not so great if those business entities end up out performing Canonical in the same enterprise or cloud servicing spaces, when Ubuntu depends intimately and exclusively on Canonical’s success. The wider Canonical casts its net looking for revenue opportunities, the harder it will be for Ubuntu to sustain and build new relationships with new business entities who will be empowered to lead Ubuntu.
    If the Ubuntu community likes what Turnkey is doing, and thinks that is the direction Ubuntu Server needs to go, then Turnkey needs to be given some control over what Ubuntu is, possibly including being allowed to control infrastructure on which the Ubuntu process is built.

    -jef

  22. March 23, 2009 at 02:35 | #22

    Whatever you do, keep the minimal install as it is.

    Those who want “cool” can go with what you are proposing.

    For the rest of us we are happy to have Ubuntu Server the way it is now: a stable Debian with a plethora of up to date applications that WE can customize the way they want, and use as little or as much as we want.

  23. March 29, 2009 at 07:40 | #23

    There are already a number of differentiators between Debian and Ubuntu. The regular and scheduled release cycle of Ubuntu actually makes a world of difference if you’re looking to leverage new technology.

    Virtualization, and LVM in particular was the catalyst for most of my shop going from Debian to Ubuntu.

    But there is still work to be done on getting technology in Ubuntu Server. I was a little shocked to find multipath completely broken in Intrepid (1, 2).

    I’d agree that a friendly server focused install via Ubiquity with some GUI system administration applications to configure systems of software is essential in getting more users to Ubuntu Server, it is essentially the technology that brings us here in the first place.

    Convictions aside, there are many choices for a web server. Many of them function quite well and their users are happy with them because they function well, once they are set up. You need to draw these users away with something. Reliability and ethos have been successful in the past, but I believe what it can do is as important as how you make it happen.

    Which is what’s more important to me. I use a preseeded d-i netboot and vm-builder to deploy to hardware and kvm. Then chef configures the software for me once I’ve written the initial recipe or found one that someone has already written.

    I suppose I’m saying I’d be disappointed if the Ubuntu Server Team focused on only applications that serve people with a couple servers. Source control for /etc is great, until you no longer configure machines individually and start keeping everything in source control centrally. A wizard for winbind is nice if you haven’t figured it out (for that release).

    Maybe we need to tick off the common problems before we can move to something like an applet for configuring and monitoring GFS clusters. More likely that’s as far as we’ll go, as I have to assume they satisfy the majority of those who would pick up Ubuntu Server.

    I’ve been seeing a lot of innovation happening elsewhere, and I’d be personally saddened if Ubuntu Server became only what everyone else has done with a nice wizard on top.

  24. December 28, 2010 at 17:07 | #24

    24]Пришли схимить, а святой и черное платье хочет скинуть.
    http://linux-planet.ru/?news=5219
    команда
    wget
    yura
    сервис прокси в ubuntu 9

  1. March 22, 2009 at 11:09 | #1

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 33 other followers