back to article SQL Server for Linux: A sign of Microsoft's weakness. Sort of

Microsoft is porting SQL Server to Linux, with a private preview available now and full availability “in mid-2017”. This is a big strategic move for the company. Microsoft’s server applications, including Active Directory, Exchange, SharePoint, SQL Server and Dynamics CRM, and the various components of System Center, have …

  1. Adam 52 Silver badge

    "Porting SQL Server to Linux increases its potential market, and will help Microsoft to compete with Oracle and to run on public clouh as AWS (Amazon Web Services) which are mainly Linux-based"

    I have no issue with most of this article, but this paragraph ignores that SQL Server has been available on AWS RDS for years so is somewhat misleading.

    1. Anonymous Coward
      Anonymous Coward

      "referring only to the “core relational database capabilities”"

      This does more sense as a loss leader - Microsoft might be planning to port a limited version of the product - probably making it free. Then when you outgrow it, you have to upgrade to Windows Server to get the full y featured version?

      1. Richard Plinston

        > This does more sense as a loss leader

        Or as vapour-ware, which worked so well through the 80s and 90s. "Hold up on that convert to [MongoDB, Neo4J, PostgreSQL HStore, MySQL, ...] project, we will have MS SQL on Linux real soon now." (repeat every few months).

      2. A Known Coward

        "This does more sense as a loss leader"

        Well this being Microsoft, I expect that once they've hooked enough marks ... err, I mean customers ... they'll announce that they will no longer be supporting the linux version and existing users will have to move to the full MS stack.

        Even if it doesn't happen, who would be stupid enough to trust Microsoft on this?

    2. A Known Coward

      "I have no issue with most of this article, but this paragraph ignores that SQL Server has been available on AWS RDS for years so is somewhat misleading."

      Yes, and it's priced accordingly high though. The company I work for have an MSSQL instance in AWS, a necessity of one of our data partners, but it's costing a small fortune compared to an identical MySQL instance. If MSSQL on linux reduces the costs to the end user for Amazon RDS then it's popularity would definitely increase.

  2. BobChip
    Holmes

    Weakness? More like hard-headed reality.

    A surprising move?, but perhaps not on reflection. Windows, as a desktop OS, only accounts for about 10% of M$ income and the desktop market is declining fast. M$ see their future as being in the cloud, which is fundamentally Linux land. So is mainframe computing. While I suspect M$ did not rush into this eagerly, it was get on board or get left behind.

    Wonder what they will port to Linux next?

    1. Lusty

      Re: Weakness? More like hard-headed reality.

      "Wonder what they will port to Linux next?"

      This has nothing to do with porting to Linux, and nothing to do with SQL Server. This is entirely down to competing with AWS and becoming the second public cloud and gaining market share. Microsoft are not stupid, they can see that in the very near future software licensing will bring them almost no money. I wouldn't be surprised to see them make all their software free and/or open source. In the meantime, they have to prove they play well with others to be a trusted cloud platform. This will allow them and AWS to become and remain the two primary public cloud providers - Google may join them but their plans are not so clear and their implementation behind in features. The future for Microsoft is as a service provider and this is a step to ensure that transition is a smooth one.

      1. P. Lee

        Re: Weakness? More like hard-headed reality.

        Indeed, this looks like a Hail Mary to discourage on-prem ms database apps disappearing into Linux-only clouds with no ms services remaining.

        I can't see any Linux people going for mssql by choice or ms bods hosting on Linux by choice. This is an "it will be easier not to port our db to Linux-cloud but public cloud is strategic" play.

        Like os2 it may just provide an easier migration path.

  3. Warm Braw

    Hail unlikely business strategist Judge Thomas Penfield Jackson

    16 years ago, Microsoft was fighting a court decision to split its OS business from its applications business. Now it's looking like a missed opportunity.

    1. Roo
      Windows

      Re: Hail unlikely business strategist Judge Thomas Penfield Jackson

      "16 years ago, Microsoft was fighting a court decision to split its OS business from its applications business. Now it's looking like a missed opportunity."

      It was said at the time it would be good medicine for MS, I think it's more of case of Bill & Ballmer lacking the vision and the stones to take a short term hit for a long term gain.

  4. Tom 7

    But, I'm guessing, not to raspbian

    or any other ARM based machines.

    1. 1Rafayal

      Re: But, I'm guessing, not to raspbian

      If it doesn't get get some sort of arm based version eventually, I would be surprised.

      1. Tom 7

        Re: But, I'm guessing, not to raspbian

        There seems to be some problem with .NET and ARM. Or rather .NET and MS and ARM.

    2. Doctor Syntax Silver badge

      Re: But, I'm guessing, not to raspbian

      Informix is already available on ARM. IBM sees it as an embedded database.

  5. jake Silver badge

    Yeah ...

    ... but who in their right mind will actually use it in a production environment?

    That's not a knock on Redmond, that's a serious question.

    1. Doctor Syntax Silver badge

      Re: Yeah ...

      "That's not a knock on Redmond, that's a serious question."

      The two are not mutually exclusive.

    2. RIBrsiq

      Re: Yeah ...

      "who in their right mind will actually use it in a production environment?"

      Immediately? No one. Or so I hope.

      Eventually? Well, that will depend on how good it actually is, won't it?

    3. Pirate Dave Silver badge
      Pirate

      Re: Yeah ...

      My thoughts exactly. Why would anyone spin up a production database running on an OS that Microsoft may decide to stop supporting or even developing for in 5 years (or less)? High and dry, that's where they'd be.

      So if nobody puts MS-SQL on Linux to "real" use for fear that MS will yank the rug out from under them, it will remain a niche player and eventually die due to lack of interest.

      http://www.aesopfables.com/cgi/aesop1.cgi?4&TheScorpionandtheFrog

      1. Sandtitz Silver badge
        Meh

        Re: Yeah ... @Pirate Dave

        "Why would anyone spin up a production database running on an OS that Microsoft may decide to stop supporting or even developing for in 5 years (or less)?"

        This all will depend on the support policy MS will publish if/when this SQL server comes to market. Microsoft has typically offered 10 years of support (+/- year or two) for each SQL version for the last 20 years. They'll probably support it only on specific SLES versions at first since SUSE and MS are in bed together.

        Speculating with the yanking of the rug - that's just FUD, man.

        1. Pirate Dave Silver badge
          Pirate

          Re: Yeah ... @Pirate Dave

          "Speculating with the yanking of the rug - that's just FUD, man."

          No, that's something we have to seriously consider when we think "hey, MS has MS-SQL on Linux now. Should we move our student records database to that platform?"

          Let's face it, the only real guarantee from MS about MS-SQL is that it will always run on Windows. Anything else is a gamble in any mid-to-long-term scenario, regardless of what comes out of the mouths of the marketing droids this week/month/year. So the safe bet, if MS-SQL is needed, is just to run it on Windows. Otherwise, in 3-5 years, it could be Zuned when Microsoft "re-aligns our core competencies".

          I mean, it's good (overall) that MS is offering MS-SQL on Linux, but at the same time, those of us who need long(ish)-term stuff are going to be wary of this (besides which, with EDU vloume pricing, a standard Windows Server license is only $50 a year or so). Internet startups will probably eat this like candy, though.

  6. Velv

    Why?

    Quite simple: Oracle

    Oracle is now enforcing its licensing stance on virtualisation, and if you virtualise your Oracle on VMWare (which is not on Oracles approved list of partitioning environments) then you need to license EVERY CPU controlled from the same VCenter, despite it probably running on only a few of them. The analogy used elsewhere was like NCP charging you for every parking space in the multi-storey car park simply because you want to park in one of them.

    I can see a substantial number of Enterprises switching away from Oracle, and they probably already have SQLServer on their approved and supported list, so why not run it on Linux. They're less likely to switch to a different Linux based DB simply because of the support agreements they already have.

    1. Anonymous Coward
      Anonymous Coward

      NOT EVERY CPU in vCenter

      The statement about having to license every CPU in your vCenter for Oracle is just wrong.

      "VMware is not recognised as an approved hardware partitioning mechanism for Oracle" - True. But this is for sub dividing multi cpu machines into smaller single/dual cpu boxes.

      You have to license all sockets in a multi socket box running Oracle on VMware. You also have to license any other clustered hosts where the Oracle VM could run due to automatic DRS, HA etc.

      Any host where Oracle does not run and cannot run without you moving it there does not need to be licensed. Otherwise you get the absurd situation where you are licensing all physical hosts where you _could_ run Oracle - and I don't just mean in the same vCenter.

      Example: All hosts are connected to the same FC SAN - I _could_ vMotionn an Oracle VM to a different host on a different vCenter in a few seconds - or maybe even a physical box with a shutdown/restart. So now we have gone from the same vCenter to all your SAN connected machines. (Hope you don't have any big iron UNIX boxes.)

      What if your SAN uses iSCSI - Now you are in real trouble because that could be every machine that talks IP. (Hope you don't have an Internet connection)

      Just read your Oracle contract - it is fairly clear on this point.

      1. Roo
        Windows

        Re: NOT EVERY CPU in vCenter

        "You have to license all sockets in a multi socket box running Oracle on VMware."

        Surely that's worse than paying for every CPU - because that would include unpopulated sockets as well. In effect Oracle would be charging you for a CPU that doesn't exist...

    2. Paul Woodhouse

      Agreed, this is basically gunning for Oracle I suspect :)... no bad thing, SQL Server is about the one MS product I do prefer using over Open Source stuff a lot of the time... not dissing mySQL or Postgre if I spent more time with them I'd prob. be happier with them, but I really do like SQL Server for some reason...

    3. Gazareth

      "then you need to license EVERY CPU controlled from the same VCenter"

      I don't think that's too different to how you have to license SQL on VMWare/other virtualisation platforms.Gets a bit fuzzier with SA though.

      Although I'm no licensing expert.

  7. Anonymous Coward
    Windows

    The competition awaits...

    I think Microsoft are in for a hard time because just like the mobile market is already more or less saturated, the same can be said for the availability of SQL environments running on Linux (and other Unix-like variants).

    Of course I'm looking at MySQL and PostgreSQL for starters (which are also available on Windows I might add) but there's a whole lot more than that. Firebird SQL, MariaDB and MongoDB (all available on both Windows & *nix) are players as well.

    And competitive issues aside; even though I actually like Microsoft SQL server (to a certain degree anyway) I can't imagine how this would work out on a *nix environment, which is mostly dominated by commandline usage. Yet one of the MS SQL server's strength is also its interface and its hooks into MMC (and Powershell of course).

    Take all of that away and I honestly can't help wonder why I should pick MS SQL server above PostgreSQL (which is my favorite pick for larger (enterprise-like) environments).

    1. Anonymous Blowhard

      Re: The competition awaits...

      "I can't imagine how this would work out on a *nix environment, which is mostly dominated by command line usage"

      I would expect that a Linux based instance of the SQL Server DB engine would be managed using SQL Server Management Console on a Windows client; although Microsoft might also port this to Linux.

      I think that part of the logic for SQL Server on Linux will be to lower the price for using SQL Server in an application; for smaller systems, where SQL Server Express will suffice, it will effectively be free.

    2. Anonymous Coward
      Anonymous Coward

      Re: The competition awaits...

      Take all of that away and I honestly can't help wonder why I should pick MS SQL server above PostgreSQL (which is my favorite pick for larger (enterprise-like) environments).

      You wouldn't, and it is worth observing that the Really Big Boys don't run Microsoft at all in production. The likes of Amazon and Google will laugh in your face if you'd propose a Microsoft product - why blow a lot of money on an as yet unproven platform when it is possible to go entirely without?

      I reckon Microsoft has an uphill climb ahead..

    3. Pirate Dave Silver badge
      Pirate

      Re: The competition awaits...

      "Yet one of the MS SQL server's strength is also its interface and its hooks into MMC (and Powershell of course)."

      MS started this with Server Core - they don't want you actually touching the server directly, they want you to do things remotely.

  8. Anonymous Coward
    Anonymous Coward

    .NET stored procedure

    Calling out to a .NET assembly is rare, or should be. (The article wording makes it sound like all stored procs are .NET to me!) If you try to put such a thing in my databases you'd better have a bloody good explanation for why it were necessary and couldn't just be done with normal TSQL.

    1. Paul Woodhouse

      Re: .NET stored procedure

      damn yeah, why can't SQL server support regex properly?, thats about the only use I have for .net assemblies in SQL server..

    2. Adam 52 Silver badge

      Re: .NET stored procedure

      I did quite a lot of benchmarking this for my one of my degrees. Calling out to .net for my use case (web log parsing) was dramatically faster than doing it in SQL and took much less development and support effort than using an external process.

      1. Anonymous Coward
        Anonymous Coward

        Re: .NET stored procedure

        Hope you failed your degree. Why the hell do you think it is the job of a database engine to parse web logs? You have an external tool parse the web logs and then insert into the database.

        1. Paul Woodhouse

          Re: .NET stored procedure

          in fairness he does say that it took less development and support effort than using an external process, guessing as it was part of a degree he would have had to have written an external process from scratch to do it... in the real world though, you'd prob. just grab an existing tool to do it...

        2. TonyJ

          Re: .NET stored procedure

          "....Hope you failed your degree..."

          And let's follow the trend of personal attacks/uncalled for nastiness being posted anonymously.

          You're not so much a prick as an entire cactus.

        3. Roo
          Windows

          Re: .NET stored procedure

          "Why the hell do you think it is the job of a database engine to parse web logs? You have an external tool parse the web logs and then insert into the database."

          All you've done is hurl some abuse and advocate a solution without first taking the time to understand the problem, I can't see any upside to your method of working. Hopefully the OP did get their degree and is enjoying a successful career in engineering without having to live with the damage done by abusive anonymous opinionated commentards.

          1. Anonymous Coward
            Anonymous Coward

            Re: .NET stored procedure

            So you think degrees should be handed out to people who present academic work that is 100% wrong from an academic standpoint? I would not want to use a system engineered by that individual, or yourself as you seem to support it.

            1. Roo
              Windows

              Re: .NET stored procedure

              "So you think degrees should be handed out to people who present academic work that is 100% wrong from an academic standpoint?"

              I don't, however an undergraduate displaying the poor social skills, poor reading comprehension and poor rhetoric that you have displayed in this thread would require that to be the case.

              "I would not want to use a system engineered by that individual,"

              Fair play, for the record I wouldn't want to use anything that you were remotely involved in either, and I believe that you would share my view if you took an honest look at your "shoot first" "I know best" attitude.

          2. aazmi615@gmail.com

            Re: .NET stored procedure

            My friend, insulting someone is a poor strategy for helping them out.

            When you give, don't take.

  9. Novex

    Perhaps this is a sign that Microsoft is realizing it has to become a software company again, rather than a closed platform system that dominates everything? Sounds obvious now I type it, but during the Gates and Ballmer eras, they just couldn't see it. However, believing that everything can be linked to cloud and controlled from there could be mis-step.

    1. Franco

      I don't think they are going for everything cloud. I know the strategy is mobile first, cloud first under Nadella but I don't see the death of on-premises Windows Server and associated products (like SQL Server) anytime soon. There is still a huge range of industry that will not use cloud services for security reasons (Companies I've worked for include Banks and Defence Contractors).

      The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first.

      1. A Non e-mouse Silver badge

        @Franco

        The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first.

        Linux servers are a much bigger market than linux desktops.

        1. Franco

          Re: @Franco

          They are, but Office is already on 2 flavours of Linux (broadly speaking, iOS and Android) and Office has been propping up Microsoft's revenues for years.

      2. TheVogon

        "The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first."

        People actually use Linux servers though. The Linux desktop market is tiny.

      3. Ken Hagan Gold badge

        "The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first."

        Perhaps they are testing the waters. Perhaps all their product teams are /considering/ this sort of move and the SQL team were asked to try it first. (I imagine that a DB engine is much easier to port than a very graphical desktop product.)

        Perhaps Active Directory will be next: either the server-side (which is another GUI-less pile of code that might be easy to port) or (more interestingly) the client-side. That is, AD support for managing Linux desktops. Quite a few people must have wondered how hard it would be to migrate "email and web" seats over to a Linux desktop. Increasing numbers are actually doing it. Providing decent AD support might be the only way MS can retain any of that business. (I note that MS have basically given up on trying to make money out of Windows client licences, so logically they shouldn't actually care about people switching to a Linux desktop as long as they continue to run MS products on the server.)

        1. Anonymous Coward
          Anonymous Coward

          "Perhaps Active Directory will be next: either the server-side (which is another GUI-less pile of code that might be easy to port)"

          No need, Samba 4 already does this very well these days for both the server side (including out-of-the-box LDAP, Kerberos and DNS) as well as the client side (winbind with nsswitch.conf and PAM).

          I can recommend this book on the subject which explains the exact situation you mention very clearly:

          https://www.packtpub.com/web-development/implementing-samba-4

      4. TonyJ

        "...The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first..."

        I'm with you on that. It seems to me that they're trying to penetrate a market (like they did with phones) that is already saturated with enough products across the whole spectrum of lousy to brilliant in a sphere they're not really trusted, so whilst the server market for Linux might be a bigger cow for them to suckle from, I'd suggest it's going to be a tough fight to get to the teat. And on top of that, they're not the big boys in this playground and can't give things away to decimate the competition.

        Even if all they started with was Outlook, I think they'd get more uptake.

        I'll be watching this with interest.

      5. Maventi

        "The only surprise for me is that it's SQL Server going to Linux first. I honestly thought they'd port the cash cow that is Office to Linux desktops first."

        Not really a surprise - they are playing their cards well here. I imagine the effort to port SQL Server to 'nix would be much easier than Office, plus SQL Server being a low-volume and high-margin product this would likely see higher RoI. It's also less risky - let's face it, not many Linux shops care much for SQL Server but it will be a handy niche for those who require it.

        OTOH, porting Office would open the floodgates to making Linux desktops viable in the enterprise. Perhaps even for home users too who are fans of Office. It would be great for users to have such choice of course but at this stage far too risky for MS - it would be the thin end of the wedge.

  10. Mike Shepherd

    I came to work this morning, so it makes no sense ever to go home.

    "Porting it to other operating systems made no sense; after all, it was itself originally a port".

    Perhaps the author could explain the logic of this.

    1. Zippy's Sausage Factory

      Re: I came to work this morning, so it makes no sense ever to go home.

      I read that bit as meaning that porting it would anger the partners whose IP they originally licenced, and it would have made no sense from a "not getting bogged down in expensive and annoying legal cases and potentially have to drop the product altogether" kind of way.

  11. Robigus
    Meh

    Been there, done that, am suspicious.

    Having worked extensively on Sybase, SQL Server and PostGreSQL, on Unix/Linux and Windows environments, I hesitantly welcome this news.

    Although the tooling for SQL Server is pretty good, I find that encourages users to engage in the 'click dance' of left click, right click, third item down, go. I consult to a multi-million pound company who had kittens when I wrote a script to move a database and it's logs onto another drive, instead of looking at the familiar SSMS interface that "Is the database".

    And there are of course the licensing costs; servers get downgraded to poverty editions whenever possible and functionality removed. Every week, a new business problem is answered by "You can do it in PostGres"

    Detaching from the tooling encourages clearer abstract thought that actually addresses the objectives. (That of course can apply to any other Visual Studio-eque "I used the wizard" scenario).

    There's a reason that my current big EU project is using PostGreSQL; we need employ noone to be the license monitor and we can sleep at night knowing that our back door won't get kicked in (fnaar) by the license police demanding spurious recompense for some arcane technical reason. That and the JSON/HSTORE datatypes and the Javascript/Python/SQL/R et al stored procs.

    1. thames

      Re: Been there, done that, am suspicious.

      With regards to scripting, the latest verbiage from Microsoft seems to be that they're dead keen on the command line via Powershell. The "click-this, click that" approach apparently doesn't scale up to large complex systems very well, so they want administrators to get used to writing scripts to automate their routine tasks.

      If I recall correctly, Microsoft recently provided a port of part of Powershell to run on Linux. I won't be surprised if they provide some more Powershell bits and bobs to handle management of MS SQL Server on Linux.

      1. Richard Plinston

        Re: Been there, done that, am suspicious.

        > Microsoft recently provided a port of part of Powershell to run on Linux.

        They provided DSC to run on Linux. As I understand it this allows Powershell running on a Windows machine to update/change the configuration of a Linux machine.

        It's a trap. It might even upgrade the machine to a 'free' Windows 10.

  12. DrXym

    So what's getting ported?

    SQL Server by itself isn't much use because so many deployments would use windows authentication, active directory, .NET and so forth. It's heavily tied to Microsoft technologies and indeed would be heavily used by software written with Microsoft technologies.

    So why are they porting it? WHAT are they porting - just SQL Server or all the things commonly used with it? And is SQL Server going to be performance optimized? Or is this just an exercise in tokenism - throwing out some half-baked port to tick a box but not expecting or wanting people to use it?

    It reminds me a little of Internet Explorer for Unix and Services for Unix where were real things too and both sent out to die.

    1. Zippy's Sausage Factory

      Re: So what's getting ported?

      .Net's already available on Linux and OS X, and they're just going to expand that. Active Directory as a service already exists. SQL has always supported its own authentication, and outside of a developer's PC you're probably not using AD authentication anyway (yet), so I don't see a problem.

      As for whether it'll run better than a three legged dog in lead boots wading through treacle, that's the killer. I doubt the port will be even half baked, but it'll be disastrous enough that the Windows team can start screaming "see, NOBODY wants SQL on Linux!" and the internal warfare at MS will start again

  13. HmmmYes

    I guess there's a bit of pragmatism involved here.

    MS cannot get a lot of people to fork for both the WSOS, which no-one is that fussed about as a service, and the DB, which people are - DB run enterprises. So, they just go for selling the DB as the $1 kiddy bag with the hope of the OS business later.

    It will also tie in with running Linux on Azure, hosted on WSOS.

  14. phuzz Silver badge
    Linux

    Active Directory on linux

    Everyone seems to have forgotten (or didn't notice when the news was announced) that v4 of Samba can act as a fully compliant Active Directory Domain Controller.

    Not to mention that AD is a variant of LDAP, so there was already the opportunity for a lot of integration between Windows and Linux networks.

    1. Ken Hagan Gold badge

      Re: Active Directory on linux

      I'm sure MS haven't forgotten that. This might be part of the response. If we ever actually see the year of Linux on the desktop, MS are totally up a gum tree unless the rest of their product line runs on Linux.

      What do people think will be Windows' market share on the client desktop in 10 years time? Yes, there are lots of Win32 apps that will run on nothing else, but with each year that passes the older apps are less important and newer apps are more portable.

      (Afterthought: and if MS *do* need to go multi-platform then it is best to start the process whilst they still have a cash cow or two to finance it.)

    2. Jeremy Allison

      Re: Active Directory on linux

      Nope - I have a lot of users who haven't forgotten that Samba4 == AD-DC. I fix bugs for them every day :-).

  15. CAPS LOCK

    I imagine this is really for off-the-shelf applications which require MS SQLS...

    .... why would you put up with all the the nonsense if you use Linux?

  16. AndyDoran

    With a strong movement towards self-contained Appliance Virtual Machines, Docker Containers, Nano Server, the importance of the OS is fast disappearing. Why would Microsoft spend money to port MSSQL to Linux, when it could simply bundle MSSQL with Nano, or a Docker Container? Applications access the database via ODBC/JDBC/SNAC/etc so they don't care what OS is running underneath.

    If Microsoft don't port the "fancy" features of MSSQL, then who would use MSSQL on Linux? Any potential customer, who is already a Linux house, will choose one of many well-established RDBMSs rather than MSSQL.

    As someone earlier mentioned, maybe this comes from their integration of R, from Revolution Analytics, into MSSQL.

    Or perhaps Microsoft reckons a Linux OS requires fewer patches and reboots...?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like