Predictions for 2012

I know it’s illegal in Canada to claim knowledge of future events but I am allowed to think and extrapolate from what’s happening now to a few months down the road, so here goes:

  • M$ will release “8” early for ARM. They have to and because there’s less lock-in on ARM (none), they can. I think panic has set in Redmond and they are extremely anxious to push something out the door before Android/Linux declares victory over Apple. I would bet the ARMed version comes out months before the x86 version. This is new territory and will increase divisions within Wintel because many of the ARMed players are outside the x86 camp. “Partners” will be feeling betrayed. This will also put another nail in the coffin of x86 desktops. If different is OK for M$ in mobile, it is OK with consumers. Welcome to Android/Linux and GNU/Linux folks on retail shelves everywhere.
  • GNU/Linux will arrive on retail shelves everywhere in every format. This is inevitable and we saw a few more devices available in 2011. 2012 will open the floodgates. M$ blinked by going to ARM. The world noticed and GNU/Linux will be able to rush in through the opened door. Retailers and OEMs will take the opportunity to make more money selling products running on FLOSS. The x86 OEMs are hungry and wanting to feed at the same trough the ARMed OEMs have been feeding at, both Android/Linux and GNU/Linux.
  • LibreOffice will overtake OpenOffice.org. The flood of factory-installed GNU/Linux systems will trigger the change. Consumers and businesses will see the advantages of PCs equipped with an office suite at no extra cost.
  • ARM will invade the server space. The products became available in 2011. They will be adopted widely in 2012 from small servers on LANs to racks and rooms and facilities full of them. It just makes sense that fanless units at lower prices will be popular and the superior power consumption will be the deciding factor in large roll-outs. Software can be ported. Power consumption cannot be lower with x86 except at much greater cost and in the future. ARM has lower power consumption now. There is no need to wait on Intel. M$ isn’t even interested in this space so it will be GNU/Linux all the way.
  • Google will win over Oracle lifting a cloud from Android/Linux. At the same time Barnes and Noble will trounce M$ lifting a cloud from Linux. This will inflict a death blow on software patents. Legislators and courts will take note of the stupidity of software patents and trash them.
  • The result of all these changes is that M$’s bottom line will take a serious hit. No growth is visible. A downward slide will begin. Once there is competition on retail shelves, M$ will have to reduce prices seriously to compete. In servers, energy consumption and space will dictate ARM will play a big role. Once GNU/Linux clients are widely found on LANs, businesses will see less need for AD and servers that run it. Forget the cloud. M$ has no monopoly there. It’s hard to put a number on the combined effect but I expect it will be of the order of $1billion in the client division and server division, about the same as the cost of paying off the OEMs producing GNU/Linux netbooks. This time no payoff is possible. M$ has no leverage on the non-x86 OEMs. Even Barnes and Noble fought back… Chuckle. With the discredit M$’s software patents will receive, the x86 OEMs will be less fearful of M$ in every way. As the x86 consumer PCs subside, x86 OEMs will look to GNU/Linux to increase margins.

All in all, it looks to be another great year for FLOSS. Consumers will finally realize they have a choice and make the choice.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology. Bookmark the permalink.

26 Responses to Predictions for 2012

  1. oldman says:

    “The simple point is datacenters are out of space Dr Loser its either build new ones or start using tech like moonshot to enable more servers in less space.”

    Or you virtualize – in our case 300+ server workloads are virtualized across 13 physical servers.

  2. Kozmcrae says:

    “And Microsoft market shares didn’t crash.”

    I would say Microsoft’s stock is crash proof. I would also say it’s Bull proof. It never seems to go anywhere, up or down. It’s been flat for over 10 years. I don’t believe Microsoft’s stock will ever rise above $30 again. There are far too many challenges ahead and so far Microsoft isn’t showing signs of dealing with them in a creative or innovative way.

  3. oiaohm says:

    Dr Loser
    –“There is a serous possibility that demand might strip supply and it not be on open market sale until some point in 2013.”

    I rest my case.–
    Note I said open market. Demand might strip supply equals a few million units going out the door per month reason why nothing left in supply for general.

    “You’re not listening, are you? Assuming your figures are correct, it still doesn’t matter. There is absolutely no conceivable “work-horse” LAMP load that would ever need this amount of processing power.”

    Were you not listening. The design allows for 288 isolated LAMP systems in a 4 U case. The 5 chip for isolating cpus is there for a reason.

    Conversational blade server fits 14 blades into a 7 U case. Lets go for practical max what is 16 cores per blade. 224 Cores total. /7 that comes to 32 core pure 1 U of space. Compared to Moonshot 288 cores per 1U of space. Or basically a factor x 9 more in the same area. So make datacenter 9 times bigger or go threw the bother of migrating sites. Also Moonshot divides it cpus down better. 14 blades can really only be proper isolated into 14 servers. Or 2 per 1 U for blade vs 72 per 1U for moonshot. Yes 36 times more for light load servers.

    The cost and time to make datacenter 9 times bigger is worth more than the migration headaches that might happen. 9 times bigger is the light form.

    “Plus, even a simple LAMP stack solution involves gobs of other software. Which needs to be ported. Now, let’s assume this porting comes for free (which it doesn’t). You still have to test it, deploy it, support it …”

    That would be a problem if items like debian did not already run on moonshot. So yes all the required software to run a LAMP stack is already ported.

    Software for moonshot was ready 12 months ago.

    Migrating most LAMP sites will be no different than restoring from backup. Web developers on Windows migrate web sites from Windows to Linux all the time. Its no worse changing archs. I use to do between power and x86 all the time. Walk in park.

    Dr Loser I guess you were thinking that moonshots power had to be released as one huge blob. Not that the thing slices down very nicely as like all good arm chips when a cpu sleeps it true does sleep.

    The simple point is datacenters are out of space Dr Loser its either build new ones or start using tech like moonshot to enable more servers in less space.

  4. Ray says:

    You did said that it was a success though…

    Link:
    http://mrpogson.com/2008/07/24/the-monopoly-will-end-in-2010/

    And Microsoft market shares didn’t crash.

  5. Kozmcrae says:

    “One can always count on our Mr. K for his well reasoned debate points”

    Sometimes one can only respond in kind @ldman. The comedian Rich Little once said, “People can’t recognize their own voice when you mimic them. He used to mimic his teacher in class and the other students would crack up. The teacher had no idea why.

    So I don’t expect you to realize when you are being insulted by me. Not every time anyway. You hardly read these comments. You hardly read your own.

  6. oe says:

    I don’t think it’s failing, I’ve seen GNU/Linux netbooks come back onto some small computer shops shelves around here now, granted its not the Big Buys or WalMart or Best Buy, but I didn’t see that even a year ago. This is x86 stuff, discounting tablets and even netbooks…word must be getting out.

  7. oldman says:

    One can always count on our Mr. K for his well reasoned debate points

  8. Kozmcrae says:

    “… The resilience of the FOSS crowd in the face of abject failure after abject failure never ceases to amaze me. Every year there’s another messiah.”

    Step back everyone… don’t make any sudden moves… don’t make any eye contact… that’s right… just keep moving slowly… slowly now… that’s right… just keep moving…..

  9. Dr Loser says:

    “There is a serous possibility that demand might strip supply and it not be on open market sale until some point in 2013.”

    I rest my case.

    “So what is the smallest space you can fit 288 Intel CPU Core i5 2500K Sandy Bridge Quad Core Processors into. No way it fits in 4U.”

    You’re not listening, are you? Assuming your figures are correct, it still doesn’t matter. There is absolutely no conceivable “work-horse” LAMP load that would ever need this amount of processing power.

    Plus, the supporting ecosystem does not exist.

    Plus, Linux is not a magic wand. You can’t just “port” it over a weekend and be done with it.

    Plus, even a simple LAMP stack solution involves gobs of other software. Which needs to be ported. Now, let’s assume this porting comes for free (which it doesn’t). You still have to test it, deploy it, support it …

    … The resilience of the FOSS crowd in the face of abject failure after abject failure never ceases to amaze me. Every year there’s another messiah. Every year the messiah from last year is quietly abandoned.

    Not once do you guys actually sit down and consider things from the perspective of people who, you know, might actually be in charge of purchasing decisions.

  10. Kozmcrae says:

    Glad to hear your business is doing just fine Dr Loser. Don’t mention doing badly, you had us worried for a while there.

  11. oiaohm says:

    Dr Loser
    “A flop would be WebOS, which HP bought for a billion or so and isn’t going to see a dime off (it’s built on a Linux kernel).”
    Hp is seeing money out of this. HP printers were known for having the worst stand alone interface known to man. I am not kidding. Mondern day HP high end printers with WebOS are very nice to use.

    So HP is already seeing money back from WebOS investment. Not exactly where you would expect.

    Dr Loser HP and Calxeda’s Moonshot ARM release is Q2 2012 you can currently pre order. Yes many of these units are already presold to partners who are currently testing out Moonshot prototype hardware. There is a serous possibility that demand might strip supply and it not be on open market sale until some point in 2013.

    “But you could probably run an entire data center on Intel CPU Core i5 2500K Sandy Bridge Quad Core Processors with a decent SAN and a bit of fibre-optics and you really wouldn’t have the faintest desire to swap for ARM.”

    Arm Moonshot scary its 1152 cores fitted into a 4 U case. The 4 U case is in fact 4 independent modules. Fully independent modules. 288 cores per module. A module can be filled with hard-drives instead of processors as well. A module can be a mixture of harddrives and processes also.

    Its a highly flexible design. Each Moonshot processor card contains 16 cores and 16 sata ports. Reason each 4 cores in its die is a sata controller and a memory controller. There is a 5 chip on the cards that control cpu to cpu talk. So each card can be split into 4 100 percent isolated system talking to a network that can also be isolated into groups. Think blade server with a lot higher density. Particularly if you decide you don’t want hard-drives just raw cpu power.

    So what is the smallest space you can fit 288 Intel CPU Core i5 2500K Sandy Bridge Quad Core Processors into. No way it fits in 4U.

    Yes the arm chips have about equal processing power to a Intel CPU Core i5 2500K. There are currently experiments with Intel atom processes but those are running into issues because they are not really designed for NUMA operation like the Arm Moonshot chips are.

    Something datacenters are running out of is floor space. Remember 7000+ new servers in data centers are brought on line each day. 2, 555,000 a year. Note that is not replacement and upgrades. You have to find space to store 2,555,000 extra servers per year with current hardware.

    Arm Moonshot higher density means that you can avoid having to build new datacenters for a while. Thinking 70 percent of web service is Linux. So those will run on arm hardware without issue.

    Also Arm Moonshot generate less heat for processing power so also save on building cooling as well.

    Yes hyper-visors was the response to the last case we run out of floor space in datacenters. That is now running out of means to assist. So now the arch has to change to fit more hardware into the same space.

  12. Dr Loser says:

    Not as far as I know, but then again I’m just a contractor. My job is to rip them off for all they’re worth.

    I must try the small thingie market some time. Gotta be worth a few brussels sprouts and enough klim to keep me satiated thru the New Year.

  13. Kozmcrae says:

    “M¢”

    That’s funny Dr. But no one around here mentioned anything about your employer doing badly… Except you that is. Is your employer doing badly?

  14. Dr Loser says:

    Incidentally, if my esteemed and benevolent employer is doing so badly in the market-place, shouldn’t you start calling it “M¢”?

  15. Dr Loser says:

    “That came true. M$ is selling only 50million licences per quarter, about seven per second when people are buying 90 million x86/amd64 PCs per quarter, 13 per second.”

    Nope, Pog, it didn’t “come true.”

    Even on your own wonky figures, seven licenses per second does not a flop make.

    A flop would be WebOS, which HP bought for a billion or so and isn’t going to see a dime off (it’s built on a Linux kernel). A flop would be Ubuntu servers, which are going nowhere. A flop would be Mac servers, for that matter. A flop would be buying MySQL for another billion when you could just fork the damn thing for free, thereby bankrupting the company and being forced to sell out to Oracle.

    See, there are flops and there are flops. And then there are successful predictions.

    Go on, quote a single one to us.

  16. Kozmcrae says:

    “So where are all these non-Windows 7 computers?

    160 million a year should show up somewhere, eh?”

    Not if you’re looking for Windows 7 computers.

  17. Clarence Moon says:

    So where are all these non-Windows 7 computers?

    160 million a year should show up somewhere, eh?

  18. Dr Loser wrote, “My, the western provinces of Canada are tres sophistique, aren’t they? Back in the real world,we use air conditioning, designed and provided by professionals. Oddly enough, that’s what 100% of server farms do.”

    Of course, I had a better solution for them: buy larger/more hard drives and virtualize seven servers onto one machine. Of course that would not do… I no longer work there but they also paid for all that power several times over by burning oil in diesel generators. The oil was bought in Montreal and shipped by barge to the Arctic. Cost was about $1/KWH so the cost of electricity would have paid for a new/better server to run the whole place. Unfortunately, my bureaucrat neither paid the electrical bills nor was able to ventilate the room so the door was left open to avoid overheating. This went on all winter when the temperature was -40ish. It doesn’t have to make sense. It’s government policy. ARM would have helped them a lot.

  19. Ray wrote, “there is the prediction about Windows 7 being a flop…”

    That came true. M$ is selling only 50million licences per wuarter, about seven per second when people are buying 90 million x86/amd64 PCs per quarter, 13 per second. The “monopoly” is getting just over half. The reason? People realize they don’t need “7” for thin clients, small cheap computers or any other kind.

  20. Ray says:

    Pog: there is the prediction about Windows 7 being a flop…

  21. Dr Loser says:

    Do I think HP doesn’t know what their customers want? Well, that’s an open question, really. I know what they think they can sell to their customers. And I understand that they will use marketing to pretend that they can deliver MagicPixieDust(TM).

    But, no, I am prepared to bet you twenty dollars that NO, HP will not ship a single ARM-based server in 2012. (The number 1 or 2 server maker on the planet!)

    Here’s the deal. I’ll send you the $20 via Paypal or whatever, because I trust your morals if not your common sense. Pick a charity, it can even be the FSF if you want, and at the end of the year you get to give them my $20 whether I win or lose, and your $20 when you (inevitably) lose.

    “A few years ago I worked at a place whose server room was so hot they had to leave the door open and run a fan in the doorway.”

    Leave the door open and run a fan in the doorway? My, the western provinces of Canada are tres sophistique, aren’t they? Back in the real world,we use air conditioning, designed and provided by professionals. Oddly enough, that’s what 100% of server farms do.

    The Gigahertz rating of the CPU is almost always an irrelevancy. You are incredibly unlikely to max out your CPU in a server farm, these days, unless you are one of the very few operations that run massively powerful computations. These would be (in no particular order) Google, Microsoft Bing, Yahoo, Facebook (possibly), and Microsoft MSN.

    You are quite likely to max out on network latency. You are reasonably likely to max on on disk latency.

    But you could probably run an entire data center on Intel CPU Core i5 2500K Sandy Bridge Quad Core Processors with a decent SAN and a bit of fibre-optics and you really wouldn’t have the faintest desire to swap for ARM.

    You want to know why?

    Because the ridiculous cost of the supporting environment, including the supporting chip-set and absolutely all the other development, not to mention the fact that your amortization is suddenly shot to hell and you’re buying totally new and unproven hardware for no reason whatsoever except that Robert Told You So …

    … because the ridiculous cost is ridiculous.

    I notice you haven’t contradicted my point on software ports, btw. Which is fine. There are only so many of your opinions that you can find bogus statistics to support.

    (I can grep for mine!)

  22. Dr Loser says:

    “What predictions have failed to materialize?”

    The burden of proof is on you, Robert. You made the predictions. You have made dozens of them.

    Pick one that materialized.

  23. Dr Loser wrote, “For the sake of discussion, however, let’s take this server-on-ARM point. It isn’t going to happen, Robert. There is no compelling commercial reason for it; it would take vast amounts of OEM R&D and investment; and nobody cares a wet-slap about power consumption in a server farm.”

    Google, Amazon, Rackspace, and M$ all care about energy consumption in their acreages of servers. They also care about how many threads/processes they can run at one time. Going to more lower-powered smaller servers is a viable option. Look at the numbers. x86 even as an Atom takes five times as much power to run a process, any process. The clockspeed of ARM is up to 2.5gHz at the top end. Who cares if x86 can do 4gHz? The ARMed setup can do 12.5gHz in parallel and likely for less capital cost. That’s the same throughput as hex x86 at 2gHz. The technologies are comparable and size, cost and power consumption do matter.

    Besides the acreages of servers there are many buildings that have totally filled their server rooms and electrical service entrances. They need lower-powered and smaller servers. A few years ago I worked at a place whose server room was so hot they had to leave the door open and run a fan in the doorway. Power consumption matters. Virtualization is one way to handle power consumption but so is ARM. ARM is now able to run virtual machines as well. A smaller ARMed server can take the place of a virtual machine as well.

    see DailyTech

    “Through these efforts, data center efficiencies are expected to reach new heights for select workloads and applications, consuming up to 89 percent less energy and 94 percent less space, while reducing overall costs up to 63 percent compared to traditional server systems.(1)”
    see HP, the number 1 or 2 server maker on the planet. Do you think they don’t know what their customers want?

  24. What predictions have failed to materialize? GNU/Linux and FLOSS are continuing to develop at a great rate and ARM seems to know no bounds so far. Honestly I had no clue that HP would open WebOS but they did after my predictions and I think WebOS being supported by the largest OEM of PCs on the planet assures its success. That will bring GNU/Linux onto every format of PC and every retail shelf space, something (the only thing) that has been holding back FLOSS for a decade. This means 2012 will be like the damned dam has burst.

  25. Dr Loser says:

    A bit of a spray-gun approach there; let’s wait and see how 2012 pans out, shall we?

    For the sake of discussion, however, let’s take this server-on-ARM point. It isn’t going to happen, Robert. There is no compelling commercial reason for it; it would take vast amounts of OEM R&D and investment; and nobody cares a wet-slap about power consumption in a server farm. Here’s a recent Canonical-based quote from the Inquirer:

    ‘While Canonical has announced ARM support in Ubuntu Server 11.10, Baker doesn’t expect many firms to have widespread ARM server deployments soon. “We recognise it is very early on for the evolution of the ARM processor for server type workloads so this is something we don’t envisage a typical business to deploy anytime soon, but it a first step on what we think will be a long term movement towards more power and efficient processing.”

    Source: The Inquirer (http://s.tt/14czk)’

    Even Canonical, who are desperate to differentiate themselves from Red Hat in the server space, can’t bring themselves to sign up fully to this fantasy in 2012.

    And to return to my earlier assertion that you should avoid talking about software development, an area in which you are staggeringly ignorant:

    “Software can be ported…”

    Yes, and pig-iron can be transmuted into gold, given enough energy.

    Have you noticed all those #ifdefs in Gnu? (I’m not picking on Gnu. They exist in Windows also, and Windows is a pretty homogeneous platform.) That mess is there to make things work on a single family of chips. Now imagine going back and doing it all over again for ARM.

    And then there’s writing compatible hardware drivers for the thing. Even headless servers need compatible hardware drivers.

    Here’s my (safe) prediction: it won’t happen by December 31st 2012. And here’s a more general prediction: it will never happen at all.

  26. Hanson says:

    Wow. If these predictions fail to materialize (once again), will you be man enough to publicly acknowledge it?

Leave a Reply