Erik Ableson 9 minute read
May 15, 2008

The Business case for Virtual OS X Server

I wrote the article “Apple in the enterprise” partially based on the rather off the cuff idea of OS X Server being sold for virtual only sales. I’ve been mulling that thought over for last little while and the more I think about it, the more interesting it becomes.

First off, you have to remember that I’m coming from a high end virtualisation perspective where I work with medium and large businesses that are in the process of moving to virtualisation solution in order to take advantage of the ability to consolidate many servers onto fewer physical machines, abstract each machine into a few files in order to simplify backups, disaster recovery and business continuity. There are a few of very mature products on the market, most notably VMware ESX and Xen.

I’ve had the opportunity to work with a number of clients that also use Macs and OS X Server in their environments and for the IT administrators, these machines are starting to pose some issues since they are now different from all of the rest. Currently, you can safely (read supported) virtualise Windows Server in all flavours from NT up to 2008, various Linux Server distributions, and even Solaris x86.

xServes were different from the others in the near past since they were more of a UNIX machine with it’s own chipset and you could categorize them with the Solaris SPARC and AIX PPC based machines. But now that xServes are Intel boxes like all of the rest of the Apple line-up, they start looking more and more like a generic 1U Intel server to the average IT administrator.

If that administrator is in the process of consolidating all of his existing x86 workloads into a virtual environment, they start asking why are these machines special? Fundamentally, they aren’t.

Under the current sales model for OS X Server you have two options: either you get it ‘free’ when purchasing an xServe or you can purchase the retail version in a 10 client ($599) or unlimited client ($999) version. Currently I would suspect that the vast majority of retail OS X Server sales go to previously installed xServes since Apple has stuck to its no upgrade license policy so to move from major version to major version, you pay full retail for the latest and greatest. Now nothing is preventing you from purchasing OS X Server and running it on just about any Mac that meets the minimum hardware requirements which aren’t difficult to meet, but I don’t think that market is terribly large. I could be wrong, but up until now I have yet to run across OS X Server on a client site that’s not already running on an xServe.

Side note: OS X Server runs perfectly acceptably on any modern iMac, even a MacBook will do in a pinch of you’re not running a big file server with loads of hosted home directories.

Now the other aspect of Apple hardware is that it tends to have a longer useful life-cycle in the enterprise. So over time this would mean that there is a substantial park of installed xServes which are getting OS upgrades more often than they are getting replaced. From this I would deduce that OS X Server generates more revenue in software sales than in hardware sales. This is the exact reverse of Apple’s current model on the desktop, where the ‘free’ OS sells hardware.

When we return to the purchasing patterns, it would seem that the bulk of OS X Server sales are to people that already have an xServe so Apple’s no longer getting any hardware sales revenue. For those odd folks who install it on something else, it’s likely going onto an existing machine. So at the most basic level, OS X Server retail sales are not generating a lot of hardware revenue in the same way that OS X on the desktop sells MacBooks and iMacs.

Clearly Apple has an interest in maintaining the xServe platform since it generates software revenue over time. That said, any OS X Server sale should result in future software revenue as users upgrade to ensure that the latest version can efficiently manage the latest version of OS X running on their desktop machines.

Virtualisation is changing the server marketplace in a huge way these days. IT Directors are looking at technical audits that show them that the bulk of their servers are only using a fraction of the computing power at their disposition, and that consolidation ratios of 10, 20 and even 40 to 1 are feasible with massive savings realised in energy, space, cooling and maintenance contracts. Many of the sites that I visit are putting into place policies that require justification for the purchase of a physical server for a given task. Virtual machines are becoming the norm in the server space. When OS X Server is proposed to take advantage of the collaboration services, unlimited mail and calendar server accounts it looks good up until they see that they need to buy an xServe. The reply is often “can’t we do this with Sharepoint?” or “There are a ton of wiki server solutions for Linux”. While none of these approaches come close to the ease of implementation, maintenance and use of the Apple solutions, they will likely win out over any solution requiring specific hardware.

On top of that, the IT Direction wants to tout how well his virtualisation solution was able to absorb all of the applications used by the company so selling xServes becomes a very very hard sell.

With Leopard Server, Apple does offer the ability to virtualise OS X Server in the EULA. The current problem is that there isn’t a product on the marketplace that can do so reliably and in a supported fashion. (I haven’t taken a poke at the Fusion 2.0 beta that they demo’d with OS X Server at the last MacWorld so that might be hidden in there). Then there is the issue that Apple stipulates that you can only run said virtual machine on Apple hardware and quite honestly, if you’re serious about large scale virtualisation, nothing that runs on the Apple platform is up to snuff. Fusion and Parallels are decent enough for small environments and simple consolidation, but lack the mature connectivity options of virtual switches, VLAN management, embedded iSCSI and Fiber Channel multipathing, live transfer of running machines to other physical hosts etc. Parallels Server is moving in this direction, but currently it’s still stuck in the single server silo mode for the moment.

If we treat OS X Server as a software only product that has little benefit to running on Apple hardware then selling a version of OS X Server designed specifically for running on VMware ESX or Xen makes sense. Apple has practically zero engineering investment since the virtual hardware platforms are normalised and don’t change very often. This is not at all like running OS X on a generic PC with god-knows-what for a SATA controller and video card. This is just like an Apple hardware platform: very limited numbers of configuration options all with standardized hardware components.

Each sale of OS X Server that goes into an enterprise is very likely to result in ongoing upgrade revenue as the tools become integrated in the enterprise workflow.

All of that said there are a few investments that Apple needs to do in order to make this work. Number one is they need a switchers guide for IT Administrators. Some serious work needs to be done on the documentation and training side of the house, much the same way as was done at the consumer level. Other things that need to be pushed out are documents on how to integrate OS X Server into Active Directory in order to leverage the existing infrastructure to use the applications built-into OS X Server, specifically the collaboration tools, and then the messaging and calendar services.

This has already started. I’ve noticed that there are now some videos on getting started with Leopard server up on the Apple site. This is a good thing. However, every IT guy I know wants, no needs, to touch and see for him/herself how this stuff works. In order to get this product into a company, people will need to test drive it themselves in their own environment to see how it fits. The show and tell approach is perfect for iPhones and iPhoto, but when you’re planning on adding a server to your infrastructure, you need a little bit more hands-on experience.

I’m seeing better acceptance for Linux these days in many sites since people can quickly and easily take it for a spin in a virtual machine and poke at it. There’s no investment required. The virtual server hosting platform is already in place so trying out something new doesn’t even require plugging in an extra monitor or keyboard. Installing an OS into a virtual machine from an ISO image is often significantly faster than installing on a physical machine. The idea of trying new stuff out is starting to really take hold in these environments, especially since the barrier to test is so incredibly low now.

The ideal solution that I can see would be to offer an trial OS X Server appliance in the ovf format. You download the image that then request a trial key from Apple. Apple gets some idea of the potential marketplace before making the definitive step to selling the product. And in the meantime they generate some buzz around the product.

Apple lowered the corporate risk for purchasing Macs with Boot Camp. “If I’m not happy, I’ll just install Windows”. Positioning an OS X Server virtual appliance as the perfect collaboration tool for workgroups (even in an Active Directory environment) would be the perfect way for Apple to take a playing card from Microsoft’s book and embrace and extend.

Personally, I think that unless Apple gets seriously into adopting the coming server virtualisation trend, they’ll be missing a great opportunity. They have all the necessary pieces in place to make this work.

Additional note: Currently the best way to integrate UNIX systems of all flavours into an Active Directory environment is Centrify’s Direct Control software. But a server licence for Linux costs as much as OS X Server so if you’re interested in offering additional services based on standards and a better management experience for the Macs that are starting to arrive in your enterprise OS X Server is an excellent investment.