You know, I think that might have worked back in the days when OSX Server was a separate product that retailed at $500 and they could have restricted cloning to true 'server' systems and collected a decent license fee on each sale.
I think that boat sailed when they reduced Server to a $30 App add-on for Lion (and the first version of that was a train wreck, too).
Therein lies one of the problems:
To do this properly, Apple would need to turn OSX
Server over to the HW company who handle it, and stop undercutting the product. y original investment in Mac servers was in the old Mac X Server 68K and Performa stuff, and then in Mac OSX Server (Rhapsody), moving onto OSX Server 10.1, and 10.2. They all cost a bundle, but in contrast to WinDoze server OS products, were bloody cheap.
You are correct about Linux though: Why run costly HW and costly SW when you can run a Linux server, with GUI and shell tools similar to OSX Server, at no SW cost, paying only for
extended enterprise support, whilst having open-source code that you can self-modify and expand?
OSX Server should have bundled something like MacPorts or other POSIX interfaces within itself to allow use of open-source tools and repositories. Access to almost all Linux software would have increased its selling points greatly. Sadly, if you want this, you have to add it, and it's hard to market a feature that isn't locked into the OS.
For example, X11, while bundled into the OS, is not integrated into Aqua. With a little work, Apple could have integrated X11 API stuff into Aqua, to allow Linux GUI software to interface seamlessly. Bloody hell, if we had sources for Aqua, we could have done that as a community, and Apple could have made money off of our ideas,while still keeping a great deal closed.
Another big hurdle is LTO support: I can't imagine a server grid without LTO drives (and maybe autoloaders). With Linux, you can do this easily, but with OSX, you have to rely on very expensive software just to add the drivers into the OS, and those may not be R/W to other OSes that rely on TAR.
In other words, even with the expensive software, cross-platform use is spotty. LTFS is not entirely reliable either, and is far less useful than proper TAR support, *and* requires LTO5. I don't feel like upgrading to LTO5 just for LTFS and better OSX support. :/
I actually liked AFP: If you were running a hive of Mac systems, it was far more reliable than SMB, and supported more operations. Out of curiosity, is there a way to integrate the newer SMB protocols into
older OSX Server releases?
I'm still running 10.5 for compatibility reasons with some software, although that bits me in the bum for support with newer tools.
In another light, I would never use Mac Minis as servers: They've no proper expansion, SAS capability, expanded Ethernet, etc.. They also aren't rack-friendly. They may make good
office networks, but not good true servers.
I run four SAS quad-channels, with SATA multiplexers into drive arrays in enclosures. I also have some XRAID stuff, that is so problematic that it's sick.
The XRAID is a key showcase piece for how Apple lost their server market: They were still using FC when the world was going SAS,but beyond that,they went with UATA drives instead of SATA via their FC bridge, which killed the design. UATA was essentially good and dead by the time Apple released the XRAID, and Apple never make a SATA conversion, or updated the firmware to support >1TB drives.
Had it used SATA initially, I think it would still be valuable today. Now you can get a 14x drive XRAID for under £200. The problem is that any long-term support on the user end is more expensive than buying an SAS adapter and a SAS/SATA enclosure. The first time you need one
new 750GB UATA drive, you have paid for the SAS card or the SAS enclosure.
It's almost impossible to find UATA drives of that size too, and in to years, you will be using refurb or used drives, with ever decreasing lifespans, unless you feel like rebuilding the mechanism in-house. I used to rebuild Apple HDDs, from the Profile and Widget, to the SC80. Those were hard-enough,much less the new drive mechs today,with tiny tolerances.
To be in the enterprise market, you need to think five years ahead, not five years behind. I frankly wouldn't be shocked if Apple went with UATA because they had an overstock of UATA mechs and saw SATA on the rise,and wanted to dump them at a huge gain. That kind of business decision is idiotic, as it irritates enterprise clients, who aren't about to buy overpriced UATA mechs from Apple,when they can buy SATA mechs at a lower price with much larger capacities.
With regard to cloud stuff, all I can say is 'Good Luck'. If you expect your datum to exist in ten years, you are playing a crap game by going with cloudware. You can only avoid a natural so many times before you seven-out.
I relied on cloud backups for my old Palm devices, and then the services started to vanish overnight. Poof, gone, no way to recover; so I learnt my lesson.
If you want real capacity, stability, reliability and backups, you want RAID 10 or RAID 51, with many drives, plus an LTO4, LTO5, or LTO6 system, storing archived tapes
off-site. For smaller files, you may want M-Disc.
If and when optical media catches up to 1TB capacity, it may become a viable solution as well. NAND will be replacing HDD units entirely in a few years, doubling incapacity every two years, whereas HDD platters are at their limit.
Thus, with 256GB NAND at the £200 price point now, and 3TB enterprise drives about the same, the HDD media makes sense today. In 2016, you can expect 512GB NAND drives for £200,and SATA to be only slightly less, and in 2018, you can expect 1TB NAND to be about £200,and the costs of SATA to
rise, due to lower production, like with PATA today.
By 2020, 2TB and 3TB NAND drives will likely entirely replace platter-based drives, except in very-high RPM needs, although NAND speeds may increase in the same curve.
NAND, by the way, is not as secure for
longevity as I would like. It has an unproven long-term shelf life, whereas M-Disc is supposedly tested to a 1,000-year lifespan ad LTO is about 35-years, expected, possibly more.
At present, in contrast, the longevity of a cloud service can't be anticipated beyond two to five years, mainly because it is a model that depends on other companies for its continuation, and any poor investment by those companies could cause them to collapse.
