Microsoft has made SQL Server for Linux. Repeat, Microsoft has made SQL Server 2016 for Linux Microsoft has ported its SQL Server software to Linux and has promised to release it in full by next year. From today, the Windows giant, which once likened Linux to cancer, will show off to a lucky few a preview of its SQL database's core engine for the open-source operating system. The full SQL Server 2016 for Linux will be …
Also... Hell just got a whole lot cooler...
Definitely. I just booked an appointment for a pierced eardrum with my GP. The pig squadron from the nearby airbase didn't just take off, they went hypersonic above a residential area. My eardrums still hurt from that....
Looking at the calendar once, twice, thrice - nope it is still not 1st of April.
This post has been deleted by its author
I was almost wetting myself , maybe like yourself.
I mean I can now ditch Postgres and/or MySQL ( okay oracle have probably sodded that up now ) and start paying some money to the Bill Gates retirement fund. I guess the poor so and so probably needs a few bob to buy another private island with airfield, private jets and a nice big harbour for a fleet of yachts.
I like to do my bit for charity, so buying SQL Server on Linux could be my bit for His Holiness William of Gates.
I don't think anyone would want his holiness to have an austere retirement.
So come on ditch postgres and start buying SQL Server on Linux !
For me, oracle because I have more experience with it. I deployed oracle instead of mssql for my vcenter dbs. All of our production shit is mysql. It works fine though instrumentation is a joke.
Maybe I'd consider mssql on linux after a 3 to 5 year track record. In the meantime any db needs to run on linux. I can do windows OK (only one on my team that can). More comfortable with linux of course though.
Scale ?
Not saying that SQL Server can't scale, but it seems to require a deal more effort where the big boys take large scale databases in their stride (which of course can lead to the reverse problem, where using them to manage smaller DB's seems like (and often is) overkill because those systems are built to deal with the really big stuff as the norm, with all the unavoidable scope and complexity in their management tools that this entails).
Depends what you mean by scale. SQL Server scales to large databases much better than either Postgres or MySQL. We have no trouble running SQL Server up to 5TB and one of our suppliers runs it to 10TB for us. In house we move to a clustered db once we get above a few TB.
There are a lot of myths about SQL Server not scaling, they tend to come from someone trying to run a big DB on inappropriate hardware, given proper hardware (i.e. not running everything off of one disc controller) it's fine.
All of our transactional Oracle stuff we've migrated to MySQL or SQL Server, the Oracle fees are just too high.
For most purposes, SQL Server scales fine. I've seen it running just fine in enterprises with something like a thousand concurrent users. Sure, that's by no means the biggest company out there, and it certainly leaves space for Oracle above it - but it is in the top 2% or thereabouts.
No, the real story here is going to be in the small print. This isn't "embrace, extend, extinguish", this is bait and switch. "We're porting SQL Server to Linux! Except if you want all the features, you'll need to switch to Windows Server."
"Oracle and DB2 scales up to System Z which is a magnitude over the max 8 socket PC/Wintel architecture."
Uhmm... NO!
http://www8.hp.com/lamerica_nsc_carib/en/products/integrity-servers/product-detail.html
That can run Windows Server. Scale out is usually the cost effective way to go these days though...
http://www8.hp.com/lamerica_nsc_carib/en/products/integrity-servers/product-detail.html
Agree, the HP Superdome X has solved, or at least mitigated, the scale up problems with Windows and x86 in general. I still say that it is a moot point, world runs on scale out and cluster. Paying a giant multiple for scale up doesn't make much sense as almost all apps cluster now... but if that is what does it for you, go ahead and run SQL on Superdome X. Got that base covered too.
"Oracle and DB2 scales up to System Z which is a magnitude over the max 8 socket PC/Wintel architecture."
Oracle is no longer supported on z/OS. You can run it on zLinux, not a popular option but possible. DB2 does scale well on z. I used to work at IBM and the vast majority of DB2, in general, runs on z. The cost of running DB2 on z/OS is high, but it works well. You can also buy Power boxes which scale up to 32 sockets, which would be the Oracle option (although IBM is getting rid of the scale up systems because of demand). Oracle dings you on software licensing cost if you use Power, but it is way better than anything they have in their Sun portfolio... or, I should say, because it is way better than anything they have.
I wouldn't argue that z is the best scale up platform out there. The issue is that scale up isn't really the way the world works to an increasing extent. Scale up began to go out when the web scale providers told Sun Micro that there was no way they were paying $x million for Unix scale up back in the day and would figure out a way to scale out and cluster PC boards instead... which they did... now you can use a 10,000 socket architecture on an Azure or Amazon cloud and the scale question is moot.
"With sufficient thrust pigs fly just fine"
The serious point is that a lot of SQL Server installs were done by people from a desktop background, who saw nothing wrong with a 500GB database with a 50MB/sec disc subsystem.
Oracle DBAs tended to be more experienced and better at specifying a server. Also Oracle would/do refuse to support inadequate hardware.
You still see the former thinking in some places, btw, with Hadoop clusters based on machines where the IO system is way underspecified for the processing power.
The serious point is that a lot of SQL Server installs were done by people from a desktop background, who saw nothing wrong with a 500GB database with a 50MB/sec disc subsystem.
Another serious point is that many SQL Server installations were done by people who thought they needed an "industry standard" RDBMS but whose database requirements were actually only for a few hundred to a few thousand records (of just a kB or two) in a single table. For that matter I've seen Oracle deployed for smaller datasets ("because we've got a lot of Oracle experience in-house, and our customers will pay for the licences if we tell them they have to").
Sometimes you really do need a huge industrial-strength DBMS, but surprisingly often SQLite will do just fine!
"With sufficient thrust pigs fly just fine"
Manouvering and landings remain problematic issues
In this context the pig is mysql. It's great for small jobs but eats systems rapidly as it scales.
Not that I'd let MS-anything near my core business after enduring their various clusterfucks (Especially their "high availability services" - but that decision was already signed off when the steaming pile of odure arrived on my desk. The support load is up a factor of 20 compared with the old system)
Anybody doubting SQL Server's ability to scale need only look at Sybase (now SAP) SQL Server (ASE) - which MS bought rather than reinventing YADBMS. The Sybase version is used by most of the financial institutions because it is fast, it scales, its low maintenance and it is reliable.
In my experience,the biggest problem with operational DBMS scaling is that many DBAs cut their teeth on Oracle and then treat all other DBMS as if they were the same, this is disastrous. CIOs, HR, recruitment agencies and "web developers" know scandalously little about DBMS, and care less.
SAP are desperate to sell Sybase DB stuff, The margins they offer resellers need to be seen to be believed. That's before you ask for more (they'll give).
I've never heard anyone from SAP mention Sybase without saying how SQL Server is 'the same engine'. They seem to forget that the split happened nearly 20 years ago.
" Sybase version is used by most of the financial institutions because it is fast, it scales, its low maintenance and it is reliable."
Come on, man. I used to work at IBM and have played this game. Some, not most, financial institutions use Sybase because it was popular when they implemented it in the 90s and it would be a real hassle to pull out at this point.
SAP went around saying "Wall St runs on Sybase" when they acquired it. Well, it doesn't (it runs on DB2 or IMS on z for the important stuff), but what large banks run on has nothing to do with what is best technologically. Banks, probably more than any other industry, are hesitant to pull, or even touch, working prod systems. Sybase was yanked in every other industry but some of it hangs around in banks for that reason.
SAP went around saying "Wall St runs on Sybase" when they acquired it. Well, it doesn't (it runs on DB2 or IMS on z for the important stuff), but what large banks run on has nothing to do with what is best technologically. Banks, probably more than any other industry, are hesitant to pull, or even touch, working prod systems. Sybase was yanked in every other industry but some of it hangs around in banks for that reason.
Sybase certainly has history with good performance, and doing that on half the hardware needed to run Oracle. Two major reasons why Oracle trumped Sybase in many industries were Oracle being better at marketing (and decisions being made my management rather than people who actually worked with the products) and Sybase licensing where they were trying to imitate Oracle in both complexity and cost (and effectively pricing themselves out of the market).
Oracle is only now starting to get close to what Sybase has with its Replication Server, when it comes to replication.
SQL Server most certainly can scale, if you have proper support from Microsoft and the magic "customised builds".
The Pheonix Partnership (TPP) produce an NHS medical record system called SystmOne which covers 30 million patients. The underlying database is an SQL Server environment of more than 700 terabytes, with an 8-terabyte primary transactional database. TPP servers handle 640 million transactions per day, peaking at about 34,700 transactions per second.
They are currently moving to an OLTP system in SQL Server 2014.
It's terrifyingly huge.
Shame the desktop front end is Java.
MS SQL has AlwaysOn clustering. IBM has PureScale clustering. There are plenty of alternatives. Oracle just FUDs them. I always thought Oracle RAC was a poor concept anyway as you still have a single point of failure at the storage level. "You can lose any of these servers and the DB runs uninterrupted", "What if you lose the storage array that this cluster is writing to?", ".... come on, that's not going to happen." RAC was probably a nice advancement in its day, but its day was ten years ago.
I agree with the previous commenter. Oracle had a hay day for years because they claimed that MS SQL just wasn't "enterprise grade" (whatever that means), so you had to spend large dollars for Oracle. Even Gartner, definitely a lagging indicator, put Microsoft ahead of Oracle in their DB Magic Quad. There is nothing Oracle can do which SQL cannot do which matters to anyone... and SQL is still less costly.
"It would just be a hassle to migrate."
If only DB clients all used a common language to talk to the server. Something with a proper standard behind it, ideally established years ago so that everyone had correct implementations by now.
(Edit: Yes it can't be *that* simple or else there wouldn't be so many people apparently stuck on the most expensive provider, but speaking as someone with no DB background it does appear odd. On paper, changing your DBMS should be no harder than changing your browser or compiler.)
Things really are changing over there.
Question is, how well will it run? There's quite a few differences between the typical POSIX-style OSes out there and how Windows does things, and Microsoft SQL Server was geared towards being high-performance, and so probably sacrificed a lot of portability in doing that as it wasn't a consideration at the time.
It shall be interesting to see how it goes, and good to see another contender in the Linux space.
The way I figure it, Microsoft finally realized that any effort at porting to another OS, would also afford another opportunity to recheck for old flaws in the software. Nothing like "a linux bod touched the code?!? Retest everything, cuz something will be wrong!" And lo and behold, there will most likely be - from years and years ago.
And so those missing features will be the ones they realize were full of poots that need to be winkled out first, just as soon as they can re-understand the code.
synergy/syndoche/schenectady!
I guess we could gauge possible performance by seeing how well WINE or Mono works. They both replicate large chunks of the Windows runtime environment. Anyone got any benchmarks lurking anywhere?
In fact, MS could do worse than put some decent effort into WINE, that'd make all their software available on Linux...
I heard rumours that Half-Life back in the day ran better under WINE on Linux in some cases than on Windows.
WINE though wouldn't be an accurate benchmark however, since it's a clean-room reverse-engineered implementation, and might produce very different results to taking an application written for Windows and re-writing parts to make it compile and run on Linux.
@Stuart Longland,
"and might produce very different results to taking an application written for Windows and re-writing parts to make it compile and run on Linux."
Of course one of the problems they'd have to contend with if they rewrote it for Linux is that they'd then have two code bases to support, maintain, upgrade. Perhaps they can get away with making the important bits of code portable. Who knows. MS haven't exactly done that well keeping Windows and Mac Office in sync and up to date. I guess that's a very different problem though - Office has a lot of GUI stuff going on.
What Else?
If MS get a taste for this kind of thing, it could start making the desktop interesting. Office for Linux? Outlook for Linux? I'm not sure anyone should inflict Visio on to poor unsuspecting Linux users...
Where might it end?! That could start being dangerous for the existence of desktop Windows. And that in turn could be a real nuisance. A Windows desktop being controlled by a set of group policies dished out from a domain controller is a very useful enterprise tool. Doing the same thing on Linux is a whole lot harder (though there are solutions out there like PowerBroker or whatever it's called now, but they're not going to penetrate through to the innards of things like a web browser). Of course, this is all wild and deeply premature speculation.
Port the Runtime - it's Less Work!
I'm wondering more and more about MS and WINE. They're tinkering a lot with iOS and Android runtimes on Windows mobile. .NET core has been open sourced and Mono is practically official. WINE is a mostly complete thing in a similar vein. MS 'finishing' WINE is certainly more imaginable than it was 5 years ago.
If MS and other companies were to start getting used to the idea of ensuring application portability by porting the runtime environments and not the applications themselves, it's going to make things very confusing!
So What About ARM
If MS have re-written MS SQL Server for Linux, does it recompile for Linux on ARM? Now that could really be interesting!
>and might produce very different results to taking an application written for Windows and re-writing parts to make it compile and run on Linux.
Granted my info is dated but what I saw from BOINC and its ilk is in general windows blew the door off *nix for scientific computing on regular non massively parallel systems (talking %10 to %50 better). Much of this was due to much more effort being put in to optimizing the compilers and application code itself for the OS running on the vast majority of desktops. I am hardly some windows lover but generally when Linux is as fast as Windows on something its a win.
> windows blew the door off *nix for scientific computing on regular non massively parallel systems (talking %10 to %50 better).
That's not our experience and we do a LOT of scientific computing.
Even using GCC instead of an optimised compiler (there are some nice ones for Linux), the Linux systems tend to be 30-50% faster as a starting point _on the same hardware_ (one of the common tendencies for MS fanbois is to compare linux on shitty hardware with windows on the latest/greatest)
Hardly some fanboi (clue is I'm not AC and post history is pretty harsh to Microsoft over all). Just going off what I remember when I was big into running BOINC on my rigs some time back. I think if you check the stats you will see Windows on like hardware racking up more BOINC points but again maybe things have changed in the last 7 years or so. Of course in general Linux especially owns scientific computing due to the scalability, open source flexibility and not having to buy 2048 windows licenses or whatever. (among many other reasons) *nix also obviously gives you a lot of advantages such as security and stability that relatively newcomers to the Windows world.
@Stuart Longland - SQL Server was originally from Sybase and ran on Unix. Sybase and Microsoft later went into a joint venture to sell a version for OS/2, and then Windows (when OS/2 fell down the stairs).
Sybase and Microsoft later parted ways, but Microsoft bought a source code license, and the rights to market the product on Windows, while Sybase retained the rights to all other markets. Sybase changed the name of Sybase SQL Server to better differentiate it from what Microsoft was selling (MS also licensed the rights to use the SQL Server name).
So it originally was a Unix product. The question is really what problems were later introduced by whatever changes MS later performed after their split with Sybase. Sybase were later of course bought by SAP.
I'm not sure whether there will be much interest in using MS SQL Server on Linux. The Sybase version has been available for years, and while it has its customers, it hasn't exactly been setting the world on fire. If someone is looking to move from Oracle or IBM (DB2), why would they pick another big proprietary vendor? It would be just jumping from one abusive relationship to another.
The only thing that MS SQL Server really had going for it is that if the customer was using Windows anyway, it let them run a single software stack from a single vendor in order to simplify support contracts. It's like why people run DB2 on IBM mainframes, or how Oracle is flogging Oracle DB on Solaris.
I suspect that this offering from Microsoft has something to do with cloud pricing, virtualisation, or micro-services in containers (e.g. Docker) rather than them thinking they can get business running it on servers in traditional enterprise Linux markets. The problem with these new markets is that people want micro (or zero) licensing costs for the OS to go along with the micro-services they are running.
Microsoft was supposed to address this market with a cut down version of Windows server, but I've heard little or nothing about it since the original announcement. It could be that this project was running into problems (possibly more to do with license enforcement than technical) and running Microsoft's stack on Linux is "plan B". This way pricing and licensing terms for running MS SQL Server on MS Windows remains as is, but customers moving to new cloud and container architectures have the option of running it on Linux rather than paying for Windows.
We'll have to see how Microsoft spins this, and then decide how much of their explanation we want to believe.
P.S. I'm waiting for our anonymous MS marketroid to reply to this spouting some B.S. about RHEL licensing costs on traditional data centre hardware, and thereby showing he hasn't a clue as to what any of the above means. Here's a hint - have a look at the specialist cloud and container distros, the ones who have oriented themselves entirely around this completely different market.
Who'd have thought it? Good old 4.2.2b being ported back to *nix after 20(ish) years. When I looked after it that code was running on at least 10 different flavours of Unix plus Netware, OS/2 and NT.
One slight correction: I believe Sybase kept the right to market an NT version and to call it Sybase SQL Server. It was only after a couple of years when "SQL Server" became identified with Microsoft that it was rebranded as ASE.
@thames,
"Sybase and Microsoft later parted ways, but Microsoft bought a source code license, and the rights to market the product on Windows, while Sybase retained the rights to all other markets. Sybase changed the name of Sybase SQL Server to better differentiate it from what Microsoft was selling (MS also licensed the rights to use the SQL Server name).
That raises the fascinating possibility that MS doesn't have a license to sell it on anything other than Windows. I wonder if the current MS management have forgotten its origins and is operating under a false assumption? I know I was.
If someone is looking to move from Oracle or IBM (DB2), why would they pick another big proprietary vendor?
Because your third party application only supports commercial databases? Having seen the licensing costs and DBA requirements for Oracle & DB2, maybe SQL comes out the least painful?
I suspect that this offering from Microsoft has something to do with cloud pricing, virtualisation, or micro-services in containers (e.g. Docker) rather than them thinking they can get business running it on servers in traditional enterprise Linux markets.
I suspect you're right. All comes down to pricing thought. If MS price it right (for traditional standalone RDBMS), they might be onto something (as there are many corporate customers who would still rather buy vendor product with support rather than use open source with paid support, most likely to ensure litigation is an option should they want to take that avenue to resolve issues).
I'd guess that the core DB stuff is basically just mmaped files and threading. Both of which have the same semantics on both systems, though there might be subtle implementation-specific differences in performance under specific circumstances / different trade-offs made etc.
Networking... well, overlapped I/O vs. epoll and friends isn't insurmountable especially on targets with SIGIO support or whatever the modern thready equivalent is.
It's worth remembering that when Sybase/Microsoft introduced SQL/Server, the only other multithreaded transactional RDBMS were DB2 (on MVS), Oracle & RDB (on VMS) and Oracle (on SunOS). Sybase on Unix & MVS(CICS region) used cooperative multithreading on a single core. It's long overdue, but standard Linux was not always mature enough for the MS low-support packaged model.
The motivation here is going after the data-grid Hadoop market that is entrenched on Linux Servers where MS can deploy PDW
Things are really changing. Satya is an impressive CEO. I think he took a bunch of Ballmer programs and just said "this doesn't make sense, guys." I wouldn't bet against Satya. He does something which is rare at large companies: He finds things which don't make sense and fixes them.
Like this situation... he understood that only supporting SQL on Windows doesn't mean people buy more Windows, it means that people who might want SQL but not Windows buy Oracle on Linux instead. If they put SQL on Linux, at least some of those Oracle users will float to SQL.... Satya has made a pretty amazing turn around at Microsoft. Not that Microsoft was ever in a desperate situation, but it was just getting less and less relevant under Ballmer every year. Now they are clearly one of the two major cloud providers and closing the gap on AWS. They have a good version of Windows (10) on the market. Taken the corporate tablet market back from Apple with Surface. Taken the email and productivity momentum away from Google with O365. They are closing the gap on Oracle in RDB. Hell, even Bing is, amazingly, picking up a little share on Google.... Pretty impressive to turn that large ship 180 degrees around in two years. Phones are still bleak, but he has a strategy in that space with universal apps, continuum and the upcoming Surface Phone where you think... "this could work"... and that is a marked improvement over the "this is never going to work" thought I had a few years ago.
Sod that, this should make Oracle sit up and think.
Oracle - You get the core DB for the license cost but everything else is usually cost option, partitioning, advanced crypto, even sodding serious backup compression is a cost option.
SQLSvr - Most of what you need is in the box, very few options are extra cost. They do this on Linux and start to muscle in on Oracle's "committment" to Linux and it should be fun to watch as Oracle RHEL starts to play up a bit when it tries to run SQL Server!
( after 17 years of working with Oracle, I think they need their cage rattled a bit more often )
"for how long is Windows going to be relevant?"
Windows isn't a big earner for them and never really has been. The money's in Office and friends - which is why Libreoffice scares them far more than Linux ever could (it's "good enough" and cheaper - the 2 data points which enabled MS to take over this market from its competitors in the 90s)
I don't think that they think it is going to be irrelevant. I think what this indicates is that they no longer think it will come to dominate the industry... and by saying "It's Windows or the highway", they are just going to push people away who might otherwise have bought some MS products, like SQL, but not the full stack. The previous Ballmer era thought was that Windows and MS stack would just take over the world, so no need to worry about Linux or anything else. Just ignore it. Satya has basically said that Linux isn't going anywhere and pretending it doesn't exist only hurts Microsoft.
Microsoft is going to sell a bunch of Red Hat by doing this, I'm sure they realize that, but they'll take that deal if it allows them to bring a bunch of Oracle onto SQL.
*supposedly* they have a skunkworks MSOffice on Linux that's been around since the Ballmer era (and predates the Android and iPhone versions that suddenly showed up). I don't think it'll ever release at this point with their focus on o365 and subscriptions.
My *general* impression though at this point, is they are looking to go ahead and kill off the legacy code thick Office and go pure subscription with the o365 "Web" Office once they get it to parity levels. This helps keep consistent subscription income going versus "ONE and DONE" licenses for Office of varying versions scattered all over everywhere which also create support nightmares for them. I *could* be wrong, but Nadella seems to be headed in that direction.
Of all the software produced by Microsoft, MS SQL is the ONE product I actually enjoy.
I'm still digesting this.. not sure how I'd feel about running MS SQL on Linux.
Wondering if they'd provide an Express version?
I have worked with Unidata, OpenEdge and MS SQL
Usually the simple stuff, definitely not an expert.
SQL Server isn't too bad. Not my favourite, but I've used it more than any other.
Although the language features are quite primitive compared to others, the query optimisation is pretty good.
Linux availability will be welcomed, as anything Windows-only a deal breaker for many... although I suspect this is vapourware, or a "lite" version.
...when Microsoft came up with it's own version of Java. It was like an attempt to hijack the standards and make everyone do things the Microsoft way and make Java theirs by force.
Now they've got their own version of Linux as well as something that runs on Linux and I'm sitting here remembering the Java situation and wondering...
Their own version of Linux might get AD integration, which is a feature of SQL Server, in that it can use AD members and groups directly in its security.
But I suspect this will be the database engine and agent. There will be no Analysis services, cut down integration services, little or no reporting service, HA groups, failover clusters etc. At least for a few years.
>>"...when Microsoft came up with it's own version of Java. It was like an attempt to hijack the standards and make everyone do things the Microsoft way and make Java theirs by force."
Are you referring to the very nice C#? Or are you mixing up Microsoft with Google because it's Google's I-Can't-Believe-It's-Not-Java language on Android that seems to really fit what you're describing. ;)
Don't forget J#: Java on .NET, which Wikipedia charitably describes as a "transitional" language.
I dunno about that, I'm sure they'd still be merrily supporting a language for the VM they failed to borg running on the completely different VM they built (or did they buy that too, I forget) subsequent to the failed attempt; if there was any money left in it (read: some poor sod caught in legacy in-house stack hell with a massive Java codebase inherited from a dead guy who wrote no documentation, with Manglement demanding everything must now run on .NET and no budget for a proper porting effort). I'm actually a bit surprised there isn't TBH, but maybe I'm too jaded.
In the immortal words of the late Admiral Ackbar:
IT'S A TRAP!
Who asked for this exactly?
Let's check; MS-SQL originally started as Sybase's "DataServer" for UNIX, bought by MS from Sybase's original Windows port, a product which is only desirable in the context of the Windows's ecosystem with its myriad of 3rd party tools and its integration with other MS tools, not free and not open source.
Microsoft can not keep Skype up-to-date on Linux or working well, the future of this "SQL Server" product on Linux is to develop "inconsistencies" that will grow bigger and bigger as time passes.
Thanks, but no thanks.
Release it in source code form on any FOSS license and maybe it would have some interest.
You do realise that Skype and SQL server might just possibly be written and maintained by different teams? It could also be possible that Microsoft might just put a teensy bit more effort into an enterprise product that sells for $$$$$ rather than a communication client they give away for free.
Why not just wait until you see the product before guessing what will happen and how it will run.
Look, Microsoft has a terrible track record in Linux, nothing that has come from Microsoft for Linux has ever been good.
The hyper-v drivers they produced were garbage until they were picked up and massively massaged into shape by external contributors.
I still remember when Linux virtual servers in hyper-v showed the drives twice with different geometries, I still remember how bad their network drivers where etc.
Heck I still remember IE 4 for Unix.
If they do not have the resources or the attitude to keep something relatively simple like Skype operating/working correctly why do you think they will do it better with a code base of the complexity of SQL server?
With SQL server there are other concerns besides the quality of the implementation, one of them is interoperability, is it really worth it to move from something like Postgres to SQL server?
What gotchas does the Linux version of SQL has? Because it surely will have some.
Are they going to port the management/instrumentation tools to Linux? I bet not and you have to use Windows.
What about the ecosystem of 3rd party tools and add-ons? are those going to be 100% compatible? I bet not, all MS stuff is intrinsically tied to a myriad of windows components.
I understand you may like SQL server and think that anything ese but MS-SQL and Oracle is worthless garbage, but things aren't going to be next-next-next done, the reality of each platform comes into play, all these gold-plated enterprise functions do not work on their own.
Committing to something like MS-SQL server only makes sense to me in a limited set of scenarios.
What about the ecosystem of 3rd party tools and add-ons? are those going to be 100% compatible? I bet not, all MS stuff is intrinsically tied to a myriad of windows components.
Why wouldn't they be? They all speak Transact SQL over TDS. I've used Sybase Open Client (and more recently FreeTDS) to talk to MS SQL Server as well as Sybase. Granted, version of TDS does matter to some extent.
Well, Skype is pretty much a law unto itself in most respects. It was bought (it appears) chiefly for the benefit of Office (the Enterprise end thereof, which is the money end). I've only once in my life met one person who paid money for SkypeOut/In services, so it's not like they were buying a cash cow.
That they continued developing Skype for Linux at all, and actually brought out a single competently-executed update to it, was a jaw-dropper to me. The MS of old simply wouldn't do that: it serves no discernible business goal, besides goodwill (the goodwill of people who aren't paying you any money, at that).
Admittedly they have dropped the ball now, because they're Microsoft and they don't know what they're doing with code they didn't design themselves from the ground up. I mean, Qt? PulseAudio? Jesus, one update that didn't utterly screw the pooch was a flat-out miracle. Unfortunately it looks like they must have laid off the person who achieved it.
Appreciate your enthusiasm and I'm all for having lots of choice in the DB market but in the big boys playground you deal with 500TB+ sized DBs, table row counts that lie in the tens of billions and single DB backup that can consume 20-25 tape/disk streams simultaneously for 8-12 hours a day, plus it has to stay running 24/7/365.
This post has been deleted by its author
It won't be open source.
"Or is it a proprietary software that requires internet access to phone home as per Windows 10, 8.1, 8.0 and 7.?."
Given that to "phone home" like those, they'd have to mandate phoning home, or indeed require internet access, no.
None of those OSes require internet access at all, not even to activate them.
Err, I mean....um....yay....and stuff.....I guess.
Seriously though, why would I use this? If I've got Linux then MariaDB and PostgreSQL are both just one command away and there's really not anything MS SQL does that's a big enough difference to justify dropping money on a license, let alone the extra aggravation of installing a product not endorsed by my disto of choice.
If you really think that SQL Server gives you nothing over MariaDB and Postgres then, no, this gives you nothing.
Agreed. Except that the comparison with MarIaDB and Postgres in the same breath is unfair. MariaDB continues the ignominious tradition of being fast but shit. Postgres just keeps getting better and better and more interesting for larger projects.
Azure demonstrates that MS knows how to scale.
But, as in many situations, this isn't really about the software at all. It's about the "eco-system": training, support, third-party products, etc. MSSQL has built up a reasonably loyal market in some areas. Some parts of this market might be looking to get with the in-crowd and Docker all the things. Early release due in mid-2017 doesn't sound that impressive but enterprise roadmaps are often at least two years out.
> Early release due in mid-2017 doesn't sound that impressive but enterprise roadmaps are often at least two years out.
This sounds like Microsoft returning to vapourware which works for them so well in the 80s and 90s. By announcing a product a year or more away they get enterprises adding it into their 'roadmaps'. In this case they could add MSSQLServer to their plans for their Linux servers alongside Oracle and PostgreSQL. If they choose MS then when it doesn't show up mid-2017 they have to switch to Windows servers to run it.
MS SQL became a good DB, *despite* the Sybase heritage. Whoever used the old versions knows how much limited it was until MS poured in a lot of development efforts.
From lock escalation that could easily lock en entire table with a few insert/updates, to a very poor T-SQL (for example, no exception), a date type that couldn't go before 1600 (I worked on a database for a museum who chose SQL Server 6.5 to run its archive, LOL!).
There's a reason Sybase ended up to be globbled up by SAP.
With a sensible implementation you NEVER have issues with multiple calendars. Just store the date in some calendar agnostic way (yes I know the baseline won't be calendar agnostic) and convert to/from the user specified calendar as required. That's what Unix systems do - storing timestamps as "seconds since epoch" and converting it to a user understandable date/time using the calendar and timezone rules currently set.
From lock escalation that could easily lock en entire table with a few insert/updates
I call BS. Lock promotion is (or was back when I was running Sybase in production) pretty configurable. If you take crap design and throw it on out of the box RDBMS installation you get what you deserve.
Just like any company of that size they have lots of departments fighting against each other. And the SQL-Server department is seeing that a whole new generation of developers isn't using MS-SQL any more. MS-SQL became popular because it was one of the servers used by the VB and Delphi community in the 1990s. The spiritual successor of those people are the web developers... and those people are almost exclusively using some unixoid system. So it's only logical for them to also offer a Linux version. They obviously don't care about the department that makes the operating system, just as that department doesn't care about the rest of the company (see Windows 8 and Windows 10).
It's nothing surprising at all. Parts of Microsoft see Windows as a sinking ship, that's normal and happens in most companies. In fact it's even healthy to diversify a bit.
Oh grow up! Name one corporation that can be trusted? We still have to deal with them all, so just bite the bullet, don't upgrade immediately when the GA comes out on any product and certainly never listen to your CTO when they say, "This new XYZ from Megacorp, we should implement it now!".
The requirements of Internet-based services for millions of concurrent users have taught MS that Windows is not up to the task.
The current market drive is all speed ahead toward the Cloud. Once there, the OS on the client side is irrelevant. All the *aaS stuff is essentially handled by Linux servers in the back end (basically nobody, not even Microsoft, runs a data center on Windows), and the youth of today is more exposed to non-Windows platforms than ever before.
Now MS is porting major parts of its portfolio to the platform. Seems to me that MS is gearing up for the day when Windows will finally become a footnote in the computing industry. There is only one important roadblock in the way before that happens : DirectX needs porting to Linux as well.
The day that happens, Windows will definitively be on the way out.
The arrival ?
For a new spec only announced in 2015 and the spec only published last month ?
Call me when it has arrived, we'll talk about then.
Only very basic applications using SQL are portable across databases. Many SQL implementation (and remember the S stands for "Structured", not "Standard") have enough differences to make most applications not portable. Beside that, often other features like transaction management are different enough to hinder easy portability.
That's why porting SQL Server to Linux makes some sense - they have to lure people into SQL Server (and avoid others migrate), because once there, it's more difficult to move out.
Indeed.
"and a well-behaving application shouldn't really care which SQL database engine it's talking to as long as it follows the standard."
Well, sure, if you go with the DHH thing about avoiding stored procedures. If you have umpteen lines of T-SQL or PL/SQL or Pg/Sql, life could be a bit harder. Pure SQL is fine for specifying states, but one usually needs to enforce rules about the transition between states, and when it gets procedural, each vendor does that its own way.
Please... Linux's happy free and open ecosystem went out the window as soon as it was picked up by for profits. Look at the Oracle vs Red Hat situation... If you think about it, Red Hat is in the wrong. Oracle, despite their poor intentions, is picking up the open code and giving people a choice on support. That should be completely acceptable in the open Linux world... but it is cutting into RH's business so they are doing everything they can to lock Oracle and anyone else out of their Linux distro. Microsoft putting SQL on Red Hat doesn't really effect the, already polluted Linux ecosystem, in the least.
I've not seen any information about it, but I'm going to guess that the Linux version will get a basic command line client, but if you want the full management suite then you have to run SQL Server Management Studio on a fully licensed copy of Windows.
Still, it'll run on desktop Windows which is cheap compared the cost of SQL Server.
>Big enterprise will always favour proprietary products for one reason offloading risk. Senior IT managers can push the risk and blame off to the vendor when things happen. You cant do that with open source.
BS, big enterprise run increasingly more of their stuff on Linux, because it simply is so much cheaper and scales much better. They are moving more and more to Linux, as we speak. Microsoft, like any other software purveyor, does not guarantee that the software is fit for any particular purpose. Do you think Heathrow gets $$$$$$ from MS when the boarding systems go titsup ? News flash: no, they don't ... MS would be bankrupt if that were the case ... remember Windows 95 BSOD'd when you sneezed ...
I trust them not. Cozy up to RH for better implementation of RH in Azure, develop tools for monitoring Linux servers (virtual or not-virtual (presumeably running under Azure)) and now announce that they are migrating MS SQL to Linux. Nope. I'm not drinking the kool-aid.
I'll stand by the rule of thumb that anything MS does they do for more to help themselves than the IT ecosystem as a whole. This leopard has not changed it's spots nor been de-fanged || de-clawed.
Rule #1: Don't do business with Microsoft.
Rule #2: See Rule #1
Some of us were burned when MS released IE on Solaris and HP-UX with the claim that we could run the same browser on all our platforms (remember Netscape charged for company use of their browser at this time). When they took the market, the releases of Unix version abruptly stopped, not even security updates.
The news that Skype is getting no love for the Linux client the other week, is the same old story.
And the statement that we "Love Linux" from the Azure team doesn't chime with this.
Now you may say that MS is a big company that Skype has nothing to do with the server division. Very true. But the Skype situation will frighten the horses with long term Linux support of MS products. So you'd think the word would go out from on high, if they are serious.
This has almost certainly been developed for web application developers/users.
The most common web server is Apache, and it's most commonly installed on Linux. But then there are a choice of databases and storage systems. Under that Apache/Linux set up nobody is really using an MS products to store their data. MS have probably come to realise just how big an arena that is. Get some of those people using MS platforms for data storage, and then have the ability to sell support services, or the ability to easily export the data into other (Microsoft) products. Another money maker.
It doesn't seem like a difficult sale.... SQL can do everything Oracle can do, more in some areas. SQL has always been enterprise grade. Windows was the issue in deploying SQL. SQL has ISV support from everyone, SQL skills are everywhere, SQL is about half the cost of Oracle... In other words, would you like to have more money or less money?... cause either of these DBs will work for you.
>>Does anybody actually WANT SQL Server on Linux?
My personal and business opinions often differ, and this is no exception:
My PERSONAL opinion: Hell no...
My BUSINESS opinion: Without that Windows Server license, there is a lot more flexibility on which pool of hypervisors to run it on, since I have a limited number of Windows DataCenter licenses to take care of X number of hypervisors. For an example - running MSSQL on Linux gives me greater flexibility in HA and balancing load as I can now float/spread MSSQL servers among 12 hypervisors instead of just 3.
They've already brought out their own distro for what runs their network in Azure. Its Debian based. At some point, its looking increasingly likely that they'll move off the current Windows Kernel and go BSD or Linux. Mac did it when they switched to BSD for their underlying OS which is helping them make huge profits, so don't know why post-Ballmer MS wouldn't follow suit. Maintaining a kernel and all that low level hardware/software crap by yourself is expensive as hell and nobody appreciates it nor is really willing to pay much for it. Much easier to collaborate on what amount nowadays to a commodity, and work on your apps, services and user experience which is where the profits are.
The Windows (NT) kernel is the _best_ part of Windows; there is absolutely no reason for such a switch.
Apple went and bought NeXT (and added some BSD stuff to it) because what they had with MacOS Classic was well... it did its job very fine at one point, but was severely showing its age with the competition having proper multitasking, memory management, etc. All of which NT was designed for and had from the start, along with proper symmetric multiprocessing (no big kernel lock - remember SunOS 4 and Linux before 2.2 or so?) and a lot of other things that are now taken for granted.
The fault with Windows lie in crappy third-party drivers, parts of the userland and supporting infrastructure (often for very good historical reasons, but making it sub-optimal for certain tasks), etc, not NTOSKRNL.
Really this is no surprise.....
M$ have already committed to m$ Linux , at least internally did people think they were going to virtualize m$ on linux just to run their own SQL database?
Far easier to port the products to linux for internal use, then just make an announcement about how much we love Linux.....
But don't get excited, it will not be a useful product. In reality, it will just be the limited free or cheap and limited package designed to do perform queries and provide a path to M$-$QL on Azure or $QL server on WinX, etc.
Not to mention extract more money from enterprise users for more incremental migration work/change.
The reality of this is already plain and obvious: Day to day tools will be lacking and there will be no capacity, portability, data and performance monitoring/management so resilience and scaling will be something you simply can't do unless you pay to give all your data to M$/Azure for the latest features. Maybe some other tools will help (ISVs like Redgate might do something) but M$ will have a strategy based on FUD, poor adherence to standards and the non-resolution of bugs that will keep them at bay.
Enterprises presently burdened by M$-$QL may choose instead to 'deploy to staging' (ie WinX not Linux or Azure) where they can sit for a few years while their IT teams deal with bugs until someone who doesn't know better decides to pay again to migrate to Azure 2.0. From there there will be nowhere to go (nowhere to hide from the bugs, the ToS, and nowhere to run...
Imagine your organisation in 5 years time; supporting Windows in production and/or gradually being consumed by the M$ cloud- will you have earned your pension for not calling the consultants to account and standing up to their plans that did everything other than move to something sustainable like PS-SQL?
mark my words.
whatever microsoft touches, they ruin.
their interest in LINUX is only so that when things go wrong, they want LINUX to be their scapegoat.
Because LINUX is open source, microsoft is free to do this. But it is a trap and the EULA needs to be strengthened to protect LINUX from microsoft.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
