Per-core licences coming to Windows Server and System Center 2016 Microsoft looks to be moving to per-core licences, rather than per-CPU licences, for Windows Server 2016. "Directions on Microsoft" chap Wes Miller tweeted links to a "Pricing and licensing FAQ" for Windows Server 2016 and System Center 2016 Standard and Datacenter Editions. Dated December 2015, the meat of the document offers …
It's that funny file called "COPYING" that appears in just about every tarball one downloads.
Often starts something like this:
GNU GENERAL PUBLIC LICENSE
Version 2, June 1991
Copyright (C) 1989, 1991 Free Software Foundation, Inc.
51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
The context makes clear we are talking about per-CPU or per-core licensing, which is nonexistent in Linux (and the BSD:s). One of the biggest reasons why cloud deployment is usually done with Linux. OS instances come and go, so tracking their hardware licenses precisely would be a pain. Seems Microsoft does not get this.
It doesn't really matter. Licensing per CPU or the like DOES exist in Linux too. If you never heard of it it's just because your horizons are very narrow -. and you use Linux just because you don't have to pay for it, probably, not for any other reason.
GPL doesn't forbid any way to sell GPL software as long as you comply with GPL. RedHat does sell its Linux per CPU - and it doesn't violate the GPL.
You may not like it, you may not buy it and use CentOS, Debian whatever you like - it doesn't change the fact it is not true that licensing per CPU is unheard of in the Linux world. And RedHat wouldn't stay in business if there were no people ready to pay to get full commercial support.
"Please will you confirm the name of the free (in every sense) version of Windows?"
Windows Mobile 10 is completely free
Windows 10 when used for small tablet devices is complete free
Windows 10 IoT is completely free.
Hyper-V Server is also completely free.
Windows 10 is NOT free. You must have a valid Windows 7/8/8.1 licence to sacrifice to enable the upgrade? to 10. Frankly Windows 7 Ultimate 64 bit is far superior to Windows 10 Pro. The cost of entry to Windows 10 is in fact very high, you must give up your old Windows licence. After 1st August 2016 it is expected that Microsoft will start charging for 10. That will speed up deployment. I have a suspicion the offer will get extended due to lack of take up. That or there will be a reorganization of the Windows group and a change of direction by upper management. A phone interface does not look good on 3 22" monitors.
RedHat doesn't license *linux* - you're required to buy a support agreement. The support agreement can be based on the CPU/core count - but anyone running any volume grabs the enterprise kits. This gets you - -- you guessed it -- access to the support facilities that RH provides. Interestingly enough, if you have a basic kit, it still runs. Completely. It complains, and you don't get updates out of the box (one can get them if one is wiley enough) but it runs without issue.
Did I ever wrote RedHat license Linux (but it does license its own additions)? Re-read what I wrote :) Still, you buy it from RedHat "by the socket". And it DOES sell Linux, GPL doesn't forbid to sell Linux. Of course it includes support - it's almost the only way to sell Linux...
For the matter, even Oracle never checked for the CPU/cores you licensed. Just, in case of an auditing, you were in trouble. Don't know Windows yet, but I won't be surprised if it would work on any CPU/core count - just, your violating the license, and if you're a sensible business, you avoid it.
Yeah, you're not understanding the licensing still my friend.
The PHYSICAL stuff is licensed. If you have Datacenter Edition (which unless you're mental you would if you were running lots of virtualisation and cloud deployment stuff), you don't license the guests, so they can "come and go".
So what happens when the cpu can have up to 256 cores depending on a hardware switch? Do I license for all of them or only what I think I might need. If I guess to low does the server start slowing down? This will make MS look like SAP, difficult to manage ans slow as sh!te. This will cause many many problems for managers while MS will claim it's your fault that your system is slow.
The new excuse from MS " You need to license more cores, afterall you did do an upgrade! "
Man I love MS banging on about "consistency".
Personally, I think per-CPU licensing is an unjustifiable rip-off as it stands because your license is to run a piece of software - if you want to run it faster by putting more grunt into the hardware then what business is it of Microsoft's? (Or anyone who does per-proc/per-core.)
You don't have to pay more for a CD if you are going to play it on a louder hi-fi system and you don't have to pay more for a DVD if you watch it on a bigger screen so why the hell should I pay more for a copy of SQL because I run it on a faster server?
I get it - the more powerful the server, the less servers you need to achieve the same performance so if I have a terminal server environment and use beefy hosts with dual 12-core procs and 256GB RAM then I can fit far more users on that than I could using single 4-core procs and 16GB of RAM.
That will result in me buying fewer licenses from MS but so what? That's one of the reasons I buy updated hardware - to consolidate my workloads and reduce license costs.
So yeah, changing to a per-core model across the board may well make it more "consistent" but it makes it more consistently a rip-off.
@Big John
Sure, but that's based on the number of people you are broadcasting the music to and the context - one could almost say the 'features' of the playback. I.e. it costs more if the music is integral to a specific class, like Zumba or whatever, but the charge is still based on the number of listeners.
Importantly, no tarif paid to APRA concerns the physical size of the gym or the facilities you have in it or how many rooms the music is playing in and certainly not how loud the music is or what equipment you are playing it back on. All that matters is the number of members you have.
I.e. it's a CAL.
APRA? Addiction Prevention and Recovery Administration (APRA) or American Professional Rodeo Association?
Using music licencing as an analogy probably isn't going to work.
In the UK we have the PRS (and also the PPL). The PRS definitely charges depending on square footage, number of classes taken, types of equipment, radio vs. jukebox vs. video player (with or without screen) etc.. Enjoy: http://www.prsformusic.com/SiteCollectionDocuments/PPS%20Tariffs/j-tariff.pdf
And a more complicated example from the PRS, that has elements of distance of hearing (so volume) for outside, as well as overlapping music played from multiple locations: http://www.prsformusic.com/SiteCollectionDocuments/PPS%20Tariffs/HR-future-Tariff.pdf
still amounts to the same: They'll try to bleed you for every penny they think they can get away with.
Edit: And the more complicated the scheme used to "calculate" a fee, the more assurance you have you're going to be shafted somewhere along the way. Obfuscation is the First Element of every scam.
@Paul
"APRA? Addiction Prevention and Recovery Administration (APRA) or American Professional Rodeo Association?"
What the hell are you talking about - are you intentionally trying to be combative? I am clearly Australian so (equally) clearly, I mean: the Australian Professional Rodeo Associate. Strewth/struth &c.
But no, for all those people unfortunate enough to live anywhere else, it stands for the "Australasian Performing Right Association". Which is kind of like the RIAA but not comprised entirely of c%$t lawyers and lobbyists. APRA do an excellent job of providing easy to understand licensing terms for business who, by-and-large, simply want to have the equivalent of the radio on in the background. They are responsive to users and, while they don't necessarily return as much value to rights holders as might be wished, they understand that the more expensive and the more complicated you make things, the less likely you are to have people complying.
Essentially, they are geared towards helping businesses play music compliantly and so they make sure that it is easy for them to do so. Having working in media and having been a member of the MEAA, I can honestly say that one of the biggest hurdles is not in intent or willingness to pay but the complication of compliance and APRA make it far easier than most. (Not that there aren't those who feel it's far from ideal.)
Yeah and in Windows land, the odds are that the bigger the server, the more users (the "number of people you broadcast the music to" (to use a ridiculously flawed analogy).
I understand the basic idea. If you only needed one license for the software itself, but could put that on a 64-core, 2TB RAM system with many TB of storage....
...it wouldn't be long before everyone did that and licensing revenue fell - and then the commercial business can't continue.
It is far too complex and costly I agree, but on the other side, I do see why they need a plan of sorts.
"Yeah and in Windows land, the odds are that the bigger the server, the more users (the "number of people you broadcast the music to" (to use a ridiculously flawed analogy)."
Erm, they have CALs to cover the number of users you have. Are you saying you should be charged twice for the same thing?
>>"Personally, I think per-CPU licensing is an unjustifiable rip-off as it stands because your license is to run a piece of software - if you want to run it faster by putting more grunt into the hardware then what business is it of Microsoft's? (Or anyone who does per-proc/per-core.)"
Suppose you ran a restaurant with an All-You-Can-Eat deal. And then one day two people came in wearing a single giant jumper and a pair of clown trousers with two legs in each half - that's the situation with multi-socket systems. Motherboards are the clothes, people are the CPUs. In the single most relevant factor (processing), it's two computers stuck together. Of course licencing models changed when multi-socket computer started to come about. If not, we'd be seeing 1,024-socket "motherboards" these days just so that people only had to pay for one Windows Server / Red Hat / Oracle licence for all their needs.
Per core is the same sort of thing. Maybe not two people in one set of clothes, more like they've started producing people with 16, 32 or 64 stomachs. They might share an oesophagus but it's just not the same customer base anymore. So you change. Especially given most other restaurants have already done so.
>>That will result in me buying fewer licenses from MS but so what? That's one of the reasons I buy updated hardware - to consolidate my workloads and reduce license costs.
Then if nothing else changed, costs per licence would rise up in proportion to the reduction in need. If Average Company needs sixteen licences of Windows Server and is willing to pay £100 per licence, then next year they only need eight, then cost of a licence becomes £200 because total value is still the same to that company. But if you DON'T shift to a per-core model, then its competitor Small Company gets messed over. They only need one licence but because average cost of licence has gone up due to the fact there's no granularity in it, they have to pay more than they need. Breaking the saleable unit into smaller, cheaper units, is a good thing because it prevents people being tied to an average that may not be appropriate for their needs.
Are any of us surprised? With the slow demise of desktop computing then the business servers are an easy target for M$ to strike at.
I wonder how many folks regret upgrading now? It's a good example of bait and switch on a grand scale. Oh, you upgraded your hardware and also took our threat/advice to upgrade your server. Here is the bad news - it will cost you more now. How's that for an immersive experience? M$ immersed in your money more like.
Oracle played the per-cpu to per-core game a while ago and there was a mighty rumpus. In the end though, the customers still had to cough up. I predict this will go the same way.
The writing was on the wall LONG ago and people I spoke to at MS back when OpsMgr 2007 R2 was released all but admitted that that was the way it would go in the future when I asked them that direct question.
It's not surprising in the slightest but it is rather annoying, even if it won't have a great impact on me, because it's just one more step in their "Cloud First" commitment. They have looked at licensing for 'cloud' and chosen the model they believe best fits that, from their perspective and then for 'consistency' back-ported that to software running on local servers.
I do understand that this consistency can make things simpler for people moving workloads around but the goal for MS is to make sure that on-site licensing is not simpler or cheaper than cloud and they are doing that, as with everything else, not by improving or discounting the cloud side but by crippling or increasing the price for onsite installations.
Of special note is that there is no configuration in which the price goes down. What a crock of s$#t. That makes this purely, 100% a money-grab. It has absolutely ZERO benefit to the customer because at BEST you pay the same.
Given that the BASE license is the same, effectively, this is not a change to the licensing structure so much as a pure add-on charge for something that you currently get as part of your license.
But it's not even a feature - there's no code here that had to be written specially to enable Microsoft software to run on more than 16 cores - that's part of it already and needed to run efficiently on more than 1 core. It's not as though something magical happens once you get to 18 cores that suddenly you have to bring some new programming to the table so there's no extra cost - in any sense - for Microsoft when your server goes from using 12 to 20 cores.
Oracle was mentioned but at least Oracle, for all their gougin doesn't have a minimum core count.
I was about to say it too.... Nadella looks more and more an Ellison-like CEO... squeeze as much money as you can from your "customers" - be it gathering and selling their data or switching to a license mode that will force you to pay more just because you change old hardware or update the OS. Heck, even 1 CPU servers need to be licensed for 2 socket with 8 cores! Probably the inSATiable hopes to move more people to his cloud - but he should be careful, moving to the cloud doesn't mean to move to an MS cloud.
Moreover, this move means many will wait to upgrade the OS. A silly move like the data slurping one in 10.
". . .but he should be careful, moving to the cloud doesn't mean to move to an MS cloud."
I don't think that matters all that much. What 'Cloud First' is really about - what it really means - it 'subscription first'.
That is what they desperately crave.
The idea that a customer could purchase a license once and then use that product long after a new version has been released seems to be offensive to them. MS, under Nadella, has the aspect of a spoilt child, insisting they get their way and that everyone must do what they say, when they say.
How dare people buy software and then continue to benefit from it without continuing to pay Microsoft for that beneficence? How disrespectful; how ungrateful!
Back to the specific change, however, both Azure and Amazon basically run as per-core licensing. It's difficult to disentangle but it seems clear that you pay licensing per core. But, while this works for smaller workloads, once you get higher, you might find that the increasing core count & power of processors means that you can do more with your Server Standard or especially Datacenter license for less.
Think about the current Xeon V3 parts, which are seeing 15% performance increase core-for-core on the previous V2 parts but now come with up to 18 cores and boast better multi-proc performance. Using the standard 2-proc license of Datacenter, that's a 50% increase in number of cores, each of which are more powerful than the last iteration - from just 18-months previous.
What that means in practice is that you may well be able to CUT YOUR MS SERVER LICENSING IN HALF. And that's just moving from the previous part - if you are running a 3-5 year refresh cycle, your older servers may will be on Sandy Bridge chips - maybe 4 or 6 core parts. You could conceivably - other resources allowing - consolidate 2 hosts, each running 4 x quad-core procs, into a single host running 2 x 16-core procs and STILL have cycles to spare. That's QUARTERING you MS licensing costs right there. Once you account for power and cooling and space, those fancy - and expensive - procs start paying for themselves.
That must be a scary proposition for Microsoft and this is their reaction to that but the their main goal is to charge per unit of usage, per unit of time. This licensing change is Microsoft trying to push that, not by making their prefered proposition inherently more attractive but by deliberately increasing the cost and attractiveness of the alternative - onsite, perpetual licenses.
And that's Microsoft right there.
According to the tweets in the story, licensing is based on the underlying hardware, not on the number of virtual CPUs assigned to the VM. So if you need to add a Windows VM instance to run a specific program, then you have to be careful about what hardware the VM is run on if you don't want to get boned by licensing issues.
Run a trivial Windows program that needs only one core on a server with 64 cores, and you still either end up having to pay for 64 core Windows licensing or else risk being labelled "pirates". I can see customers jumping for joy over this one.
This post has been deleted by its author
This is standard for pretty much everyone these days. Oracle, Microsoft, IBM, RedHat (for Jboss etc rather than RHEL) etc all use the same model
I do a lot of work with per core licensed software.
Run a VM on a 64 core host using 2 cores. Pay for 64 cores.
Run 2 VMs on a 64 core host using 4 cores. Pay for 64 cores
A rip off but if you're smart and have a big enough business requirement you can run 30 VMs on a 64 core host using "120" virtual cores and still only pay for the 64 cores.
However:
Run 2 VMs on 2 x 64 core cluster using 4 cores. Pay for 128 cores even if both VMs are on the same host!
This is apparently because it is a cluster and there 'could' be a situation where each VM is on a separate node.
This means that we end up running many different clusters with each cluster dedicated to a specific licensed product. We cram in as much as we can because as soon as we add a new host to the cluster we have to pay for the full underlying hardware used which is often over 500k per server.
Without weighing in on the perennial Linux vs Microsoft cat fight, simply saying "just use X" is largely unhelpful. Remember that this is an article about changes to Microsoft and (most) people are commenting based on their experience and foreseen future with that same company.
You do people a disservice by assuming they are not aware of or have not investigated the alternatives.
>Start clustering the VM hosts and you may need to licence for all possible hosts it could be on.
This is the headache with IBM Parallel Sysplex, licensing costs can quickly get into stupid figures.
But at least IBM did provide a means for locking/unlocking processors, so that you could purchase a fully populated box and limit the number of available processors and change the limit if you needed to shift loads from other systems.
I ended up spending 30K to rip out existing 12 Core CPUs to replace them with 6 Core CPUs (of a higher clockrate) for my Dev environment.
Ended up saving close to 1 Mil in licence costs.
There do seem to be a lot more CPUs coming through that run less cores but at higher speeds, suspect the chipmakers are realizing how people are trying to scale back on their licence costs.
(I think I got this right)
The article says: Physical cores.
so you pay for 64 cores and then your VM's are covered.
(it did say 2 VM's free, so I guess more you have to start coughing up more)
which ever way you cut it, its a con for M$ to bulk up there revenue.!
and the only reason they are doing it is because they are in decline.
they need to break up the group,
compartmentalise there ranges etc,
and realise the huge entity called M$ cant continue as it is.
as they are trying to bulk up a group, where ranges are in decline.
if they keep it up the group will start to fail.!
Every time I hear Microsoft doing stuff like this, I smile. Not for me, but for my colleagues still stuck in the trenches. Still stuck fighting the evil empire that is Microsoft. News like this just makes it easier to argue for changing the operating system, unless you're dealing with utterly clueless management. In which case I also smile when I see these companies die one by one as they get shot in their I.T. guts and have their blood sucked out by Microsoft.
<rant> I just wish non-American governments would get off the pot and refuse to use this crapware made by an American corporation, that only benefits American interests. That they continue to use it would seem to me to be prima facie evidence of treasonous conduct. ie: acting against the interests of said government. </rant>
"News like this just makes it easier to argue for changing the operating system, unless you're dealing with utterly clueless management."
Hmm, have you ever managed a large project converting 1000+ users running 30 applications, some of which are niche and rely on MS products from Windows (desktop and server) over to pure FOSS?
No-one wants to pay the ever increasing and restrictive licensing fees but maybe the management are a bit more clueful than you when it come to considering TCO, ROI, Change management, risk management, staff specialisation etc
I fail to see how being PM of a large project is necessary to understand the difficulties involved. Anyone involved in such a project understands the difficulties, probably better than the PM in the first place. I've been involved with much larger projects than the one you claim is "large". I'd consider 1000 users and 30 apps to be a "medirum" sized project, given some of the mainframe to distributed server work I was involved in in the 80's and 90's. Similar issues to the smaller "Microsoft to something reasonable" conversions I was involved in later. I was more of a tech weenie. You know, the guy the PM and others turn to when they need a solution, or need a good estimate of the time something will take, or need any number of other bits of information. I let the PM handle the paperwork though, that's what they're for really.
However, in my rather measly 32 years of experience in the field before I quit I found that it's a very rare suit who isn't totally focussed on their bonus, and little else. Sadly, bonuses seem to be more easily achieved by actions that are harmful to the company. Never understood that. I've always blamed it on the ties being too tight.
I've also yet to meet cluefull management that couldn't understand that avoiding (or getting out from if they made that mistake) being under the thumb of a single, provably evil vendor that's holding you by the short and curlies wasn't a very desirable thing. I also found that, almost invariably (I can think of only one exception, which surprised the hell out of me at the time), an honest ROI and TCO (as opposed to the lies created by the sales reps) could be written to support that position, and I wrote my share of those. Yet they were often ignored by the clueless management that kept believing the sales rep instead of the people they hired to be on their side. I think it's something to do with the upper levels of management being populated with people who mainly came up through sales. That or their ties were too tight.
"...I was involved in in the 80's and 90's. "
I think that may be your problem right there - things have changed, workplaces have changed. "Back in my day" just doesn't cut it anymore.
When you've converted a decent sized site with business critical MS based applications over to a fully FOSS environment then write it up, many people would be pleased to hear it and how easy it all was and how the costs were insignificant compared to standard proprietary licensing. If you are left with just one MS server that each member of staff needs then you still need to license all the CALs and server software. Good luck with the smooth transition of the database using vendor specific features running to to terabytes of data and multiple applications, all needing to be transferred as near to live as possible. Also good luck with your new IT team who have expertise in the new FOSS solution who know nothing about the old systems, set up and business practices working closely with the old IT team who know the old systems inside out and are soon to be made redundant.
Start from a clean slate, new startup, with a new team etc no problem. Attempt to rip the systems from under the users, customers and IT staff and then see how easy it is.
@AC writes: "I think that may be your problem right there - things have changed, "
Not as much as you and people like you would like to think. Yes, the technology has changed. Much of the background and underlying principles, however, have not. You and people like you as so focussed on the buzzwords you've lost sight of the value of experience. Which is why you casually mention how those who took care of the old systems are soon to be made redundant. I guess that's one way to make your ROI look better - fire the experienced staff and hire a bunch of buzzword compliant newbies who not only cost less, but won't have the experience to tell if what you've sold them is a pile of shite.
Sadly, your kind of short-sighted attitude is probably why so many modern projects fail. Because lessons are NOT being learned, replaced instead with buzzword bingo spewing salespeople and an attitude that anyone with experience obviously doesn't know what they're talking about.
Angry? Bitter? Fuck yeah. It's asshats like you that made up a large part of my decision (and many of my colleagues) to give up on the industry. We'd prove that most of the experienced staff (not all, there's always a few who let their brain ossify, but not that many, not in this business!) could easily learn new technology and that their accumulated experience in how the business worked was invaluable. We'd often have the project well in hand and moving along. Then buzzword spouting idiots with zero background in real "large" systems (claiming EXACTLY the same bullshit as what you've just spouted about "things have changed" when they in fact have not changed that much - it's the same bullshit just with different names, but you wouldn't recognize it because you don't HAVE that experience) but good sales skills would come in, get anyone with a clue made redundant or worse completely hostile, then blame anyone but themselves when the project failed.
That "good luck with your new IT team... who know nothing about the old systems"? THAT is the mistake you keep making. THAT is why I left, because asshats like you kept convincing clueless, short-sighted, bonus-focussed management that it was the "right choice" because it helped "reduce costs" immediately - never mind the long term cost of the company going bust because the fuckers threw out anyone who knew where the bodies were buried, or why some processes were necessary.
New technology is much easier to learn than the intricacies of the existing business systems, processes, and historical reasons for doing things, especially in a larger organization. Something people like you, who casually dismiss anyone with experience, will never understand.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
