This article is more than 1 year old
Pay to play: The hidden cost of software defined everything
Enter credit card details if you want that system you bought to actually be useful
An oscilloscope manufacturer this month slapped a hacking enthusiast's website with a DMCA take-down order after he revealed how you could unlock extra, low-cost options in one of the vendor's systems using a series of simple workarounds.
Go ahead, chuckle – it’s an obscure dust-up in a lost corner of tech, but haven’t you noticed computing is slowly going past the point of buying discrete add on physical units?
Unlocking additional functionality used to be restricted to game play, but now it’s spreading into more mainstream IT applications.
Take the poster child for cheap and open computing, Raspberry Pi, which is selling third-party codec licences to enable playback of certain video types in hardware – admittedly, the codes are very cheap indeed and not essential for what most people want to do with their device.
Bought the hardware? That's just the beginning...
The thing that is enabling makers of tech to withhold features on pain of payment is software, and with more appliances and devices becoming software defined, the chatter at this week’s VMworld, we are sleepwalking into a world were buying a device is only the first step on an endless treadmill of paying to simply turn on more features.
In other words: Pay to play, for everything. No exceptions.
It has been extensively argued that the producers of the latest generation of games consoles were essentially flogging subsidised hardware system to enable the sales of new games and media, cutting out the high street and online retailers, keeping more profits for themselves, everyone else be damned.
What is perhaps worse is that your average Joe and Jane are not seeing what is happening. As an example, a few days ago I was in a store and if you bought a certain game on release date you got “faster 9mm reloads” and “an alternate skin”. At the end of the day this is only a software key that unlocks the extra functionality that had already been cooked into the software. Activation of the said key has no additional per unit cost for the manufacturer/developer and spurs on additional pre-sales. As we all know, software components have no real duplication cost once the software or virtual item is created.
Pay to play isn't a new phenomenon; indeed, while today’s use cases might seem consumery or relatively lightweight, pay to play is rooted in solid enterprise tech.
Remember Big Blue's golden screwdriver?
It goes back through generations to the IBM “golden-screwdriver” upgrades. Golden-screwdriver upgrades were a practice on old IBM hardware where the computing giant would initially ship a larger than required system to a customer with additional RAM and CPU capacity.
If the customer wanted to activate the additional capacity at a later date, the customer would pay the fee and an IBM engineer would turn up and literally remove a screw, or in later hardware, install a software patch and the upgrade was complete. Because it was just a pure profit exercise, those upgrades were referred to as golden screwdriver upgrades. This was common practice amongst supercomputer manufacturers of the day.
As CPU cycles became dirt cheap and we moved in the era of a single server having potentially in excess of 30GHz of CPU power, the software defined concept became more and more compelling to manufacturers because it avoided hardware, returns or shipping costs.
Manufacturers jumped on and all over the wagon. Take Hewlett-Packard.
If you splash out more than $40,000 on a single new high-end ProLiant x64 server, for example, you might then have to spend an additional several hundred dollars to turn on functionality and features that are missing out of the (virtual) box.
To get full ILO (Integrated Lights Out), out of band management functionality, a customer would need to buy an ILO advanced licence, the current retail price being around $300 before tax, per server. Without this software key, you won't get items such as remote console interaction. Many a server admin has cursed HP when a server is half dead and they are unable to get console to fix the issue.
Another aspect of HP's pay-to-play plan is that on most software defined RAID setups (predominantly on low end servers), you have to pay extra to be able to use RAID 5. This can add hundreds of dollars to the purchase price of an already expensive server. To be clear, HP is not the only manufacturer doing this by a long mile, but it is a case your writer is more than familiar with, using HP equipment daily.
Storage manufacturers have not been slow to cotton on to this practice. A lot of the NAS makers use a custom operating system. This means that adding additional storage costs you extra. Until your owner’s licence is upgraded, that additional disk capacity goes unused. You won’t find that in the pre-sales marketing material.
The big question is where did it start going wrong – from the point of view of the customer – or right – if you are in the tech company’s marketing department.
To the manufacturers, it is, in essence, a way to get a device into a customer’s hands, at or just above cost, in the knowledge that at some point they can upgrade and unlock the full potential of the software or hardware and results in additional sales for virtually zero effort.
It also solves the problem of a customer being scared off by the initial price of a more highly specified system. A sales person would argue that getting their devices and hardware into a customer site is the hardest part of the job. Once you’re in, things get easier.
Sometimes pay to play works in your favour
Some manufacturers have chosen to embrace the golden screwdriver scenario but on more customer-friendly terms, including the big boy on the storage block, EMC. Larger customers can now have “bricks” of storage delivered and configured on premises ahead of time but not billed or paid for until the bricks are activated for use. Obviously a customer is expected to actually use it at some point in the near future or face stiff penalties. Brick sizes start at 30TB for those interested.
Every major manufacturer is now claiming that their software-defined switches and software-defined hardware will revolutionise the world of switches, load balancing and all manner of software-defined things. In effect, this means offloading the complexity of last gen ASICs (Application specific Integrated Circuits) into software that manages relatively inexpensive and dumb hardware.
However, I predict that all the goodies will be on a pay-to-play scale from relatively cheap to horrendously expensive. It is easy to foresee a time when software-defined switches and capabilities will be licensed by virtual ports, all managed by a licensing server that reports back what is activated at the customer site. This won't be in the near future but over time, slowly. It’ll be like boiling a frog by slowly upping the temperature so they don't notice until it is too late, with the customer in the role of the frog.
This trend will also lead to an “a la carte” menu from which a system administrator (or God forbid, a bean counter) can chose the features they want to pay for and use.
Software-defined X, or anything as a service, is here to stay and with it software-defined add-ons. It is something we are going to have to live with because for the companies making these products it is easy money at next to zero additional cost beyond providing the initial item to you.
Tech companies have a right to find new ways to make money but the onus is on them to do so responsibility and fairly. To intentionally make a product less useful or less powerful than it would ordinarily be just for the sake of making some extra revenue is bad business.
As informed buyers we need to understand that this practice exists and make intelligent decisions based on the fact that buying the hardware is now just the start of a long paying process and – where possible – source manufacturers who are a little more giving. ®