HP to begin charging for firmware updates and service packs for servers
zdnet.com
zdnet.com
So perhaps a few extra sales from people who really need the bug fixes. But at the expense of at least a loss of goodwill and increased hassle for legitimate customers.
I believe this is more widespread than you might think – many places are looking to cut budgets and if you have in-house staff who handle most of the work support contracts are a glaring line on your budget. It's not just about the customers who did drop support but also those who might be considering doing so in the future.
> So perhaps a few extra sales from people who really need the bug fixes. But at the expense of at least a loss of goodwill and increased hassle for legitimate customers.
Agreed: this is classic management-by-spreadsheet encouraging short-term thinking. I remember a few microcomputer and NAS vendors trying the same thing before going bankrupt, selling the business, etc. because everyone involved recognizes it as a shakedown operation. At multiple employers we switched to more ethical vendors specifically in response to moves like this.
Whether or not you buy another HP server is an interesting question. Dunno; this is kind of a dick move on HP's part, and not a great sign.
Or I could be completely wrong and HP is trying to boil the frog in the same way banks do with their fee structures. Either way, it doesn't seem like a smart move in the absence of a compelling "if we buy server boxes, they better be HP" story.
I really hope this is the case, and that a sustainable update model means I can reasonably expect multi-decade lifetimes for utility grade systems. My main concerns would be lock-in, and process changes that artificially create a need for continuous updates.
That's the thing, though, today, I could give him 1/10th of a server that uses 1/10th the power, and he'd have more compute resources.
That's the primary thing that kills old servers; It's not that they die (They become less reliable, but if it's good stuff and it's treated well, it's still 'good enough' for people willing to use "the cloud") What kills old servers is that newer servers use dramatically less power per unit of work done.
(I mean, there are lots of problems with the idea; it's not getting past the purest of fantasy stage, but it was fun to think about.)
The newer stuff tends to be denser and easier to manage, which further reduces rack space and meatware cost.
It hardly impacts the customers these big corporations focus on (mostly other big corporations).
I won't go into all the arguments pro and con this change but, suffice it to say, companies that have HPUX/AIX/Solaris/etc servers and cannot afford a support contract anymore are probably better served by other offerings (Linux, BSD, etc) running on less expensive hardware.
People will probably say "HP server division will die now because of this". While it's likely HP/Oracle/IBM's server divisions are not cash cows anymore, this is unlikely to happen. Just as with mainframe, it's part of IT optimization and using the right tool for the right job. Mainframes aren't gone, just hidden away in very specific areas. These proprietary Unix behemoths will go the same away... and the hidden areas they will be used at can afford the support contracts.
Overall, nothing new happening here. We are just going to repeat all the arguments again for why it's such a good/bad idea.
http://en.wikipedia.org/wiki/Linux_on_System_z
> just hidden away in very specific areas
Like every single Financial Institutions on the globe...
Horses for Courses... asking a steeplechaser to do a flat race would be foolish, but each is exceptional for the task they have been bread for.
Perhaps I should have phrased it "RISC machines" instead of specifying the OS they run.
One fault in my argument is that HP sells smaller x86 machines, not only RISC behemoths. So the small customer that bought an HP server a while ago, ran off the warranty period and cannot afford a support contract, is royally screwed. Even here, this go back to my argument: these people are not who HP/IBM/Oracle are focusing on so they couldn't care less.
Even most of the Proliant range comes with 3yrs service/support/warranty so its only really the microservers which are really affected.
Also, you are getting support contracts. Typical server timeline is paid support for 3 years, production for 5. In between you'll be adding or migrating to newer servers anyway.
Some firmware fixes are for serious thermal issues or other problems which would result in data loss.
Not only can they publish torrents targeted at the sysadmins willing to pirate the updates, they can also setup fake websites that look legitimate and offer them for free.
Especially easy considering the mess that HP support URL are.
Not hard to dump the BIOS from a known-good board and flash it into another; it's the usual way to recover from a flashing operation that went seriously wrong, and any decent repair shop will have the reader/writer hardware for this.
Specifically, I had a sun workstation (dual socket AMD) and it was originally advertised as quad-core compatible. I had a support contract for it (it was < $100). The firmware to make it support quad-core AMDs never came.
Word was that Foxconn (who actually built the motherboard) wasn't happy with the poor sales, and didn't want to spend the money on more firmware engineering. I donno how much of that was true, but it left me with a 60 pound box that could only hold four cores (and no power management).
We'll just make it barely function and the poor sap who doesn't notice any problems untill the waranty runs out will just have to pay for support.
Wonder if you could sue or ask for a refund if the hardware doesn't work and it's clearly their fault.
Aguably they were selling defective hardware that they refuse to fix from now on. That would be a fun argument in court.
Intel confirmed some time ago that HP is on the decline, while SuperMicro and others are on the rise:
www.wired.com/wiredenterprise/2012/09/29853/
http://www.softpanorama.org/Hardware/HP/ILO/#Fiasco:_some_ad...
Also we are talking about servers here and not a desktop OS.
So what's different? Everything, we are not even talking about the same thing.
Of course, no analogy is perfect, but those are the commonalities I see.
> In both cases it is software which is specific to hardware sold by the same entity which made the software.
The only thing that makes OSX Apple hardware-specific is that Apple put a check to try to prevent it from being installed on hardware they didn't produce. I can install OSX on non-Apple hardware just fine via hacks. I can also install OSX 10.9 on any Intel Mac that has the horsepower to support it.
Firmware is specific to a set of revisions of a specific board. I can't use N40L mainboard firmware on an N54L and vice versa. I may not even be able to use the same firmware on every board inside those lines.
> the software is necessary for the hardware to operate correctly
OSX is absolutely not required for Apple hardware to function correctly. I can install Linux and Windows on any Intel Mac, and Linux on any PPC Mac.
> Of course, no analogy is perfect, but those are the commonalities I see.
The analogy is shit. I can find commonalities in blue whales and elephants, but they are very different creatures. Firmware is a set of instructions for a special-purpose controller in a specific circuit. Software is a set of instructions for a general-purpose computer of a certain architecture.
However this:
> That’s a hefty price to pay to fix what is arguably a defect in the original product
What a blatant lie. How can it be a defect in the original product for an operating system which is brought out months after release of the hardware.
One interesting case many years ago was a particular motherboard that had broken ACPI support only in Linux so suspend wouldn't work. The motherboard would explicitly look for if the computer was running Linux and if it was it would hand it a mangled ACPI table that didn't work. The fix was to actually delete any code specific to Linux and have it give the Linux kernel the same ACPI table it would give to Windows and it suddenly worked perfectly. Obviously that's a problem with the BIOS and even though it said it implemented ACPI it didn't implement it properly. It worked fine for Windows but not Linux. You can certainly get more obscure problems going from one Windows version to another as well.