IBM Locks Down Licensed Internal Code On Power, Mainframe Systems

February 25, 2013 Timothy Prickett Morgan

Here we go again. Maybe. Or maybe not, if IBM‘s lawyers help it close the barn door before the horse escapes. Big Blue is tightening down the licensing of the machine code, often called licensed internal code or microcode, for selected high-end servers based on Power and System z processors. IBM has revised the terms and conditions to machine code on these machines, making it not only explicit that licenses to machine code cannot change hands, but that they may not do so without a customer signing on the dotted line a license acceptance agreement.

IBM announced the new microcode paperwork requirements in announcement letter 113-027, and they will go into effect on August 1 of this year. In that announcement, IBM said that it tweaked the license agreement for microcode on April 2, 2012, and the changes were made to clarify the terms regarding license to “machine code,” specifically relating to the non-transference of licenses between parties and the fact that only IBM can make such a license to a user. Interestingly, explains IBM: For many IBM Machines, the user conveys its acceptance of the license simply through the use of the Machine ( IBM designates such IBM Machines as “Acceptance-By-Use Machines”); for all other IBM Machines (as specified below), the new user must document its acceptance (in writing with IBM ) of the terms of the License Agreement.

The mystery is why every machine made by IBM that contains microcode is not covered by such a blanket statement, it being a pretty all-inclusive statement but for whatever reason, for certain machines, IBM thinks these terms are insufficient and is adding some belts and suspenders to protect not only its intellectual property, but more importantly against the unauthorized use of some feature or capacity in the machines. Take a look at the list of machines and their product numbers that the new licensing terms cover:

System z10 EC (2097)
System z10 BC (2098)
System zEnterprise 196 (2817)
System zEnterprise 114 (2818)
System zEnterprise Blade Center Expansion (2458)
System zEnterprise EC12 (2827)
Power Systems 750+ and 760+ (9109)
Power Systems 770 and 770+ (9117)
Power Systems 795 (9119)
Power Systems 780 and 780+ (9179)
Storage Systems 2421, 2422, 2423, 2424, 3592, 3952, 3956, and 3957

The Storage Systems above are all based on Power7 servers, in case you didn’t know that. So what do all of these machines have in common? Well for one thing, they all support capacity on demand upgrades for memory and processor cores. So it could be that IBM has caught wind that someone else has come up with a CPU or memory governor buster for either its mainframes and Power Systems and it wants to lock the licenses with users down because it has figured out that there is no way to absolutely lock the machines down from being fiddled with. It could be that Big Blue has not heard of any such thing but is just guarding against the possibility, given the relatively high prices it charges for compute and memory capacity on these boxes relative to alternatives in the market.

There are other possibilities as well. The Power7+ chip has what IBM calls accelerators on the processors that do a number of things in hardware, such as generating random numbers, running hashing and encryption algorithms, or supporting IBM’s own main memory compression algorithm, called Active Memory Expansion. At the moment, Active Memory Expansion is only available on Power7 and Power7+ machines running AIX–it runs in software on the CPU with Power7 machines and on the accelerators on Power7+ machines–and it could be that IBM is worried someone will tickle the microcode and figure out how to activate Active Memory Expansion without paying for the feature, which costs $2,600 on machines in the Power 750 and 760 class and $6,900 on machines in the Power 770 and 780 class. (You only pay one feature per server.) IBM has had memory compression on the System z mainframes for four generations now, and I am not certain what Big Blue charges for this because it doesn’t publish pricing for mainframes.

There are other funnies where tickling the microcode might get customers some feature that IBM says they are not entitled to. Just as a more recent example, in last week’s issue, I told you about a native driver set for IBM i V6.1.1 running atop Power7+ servers that is not part of the normal IBM i 6.1.1 distribution. Or, maybe it actually is in there and you just need to figure out how to act like this new software feature, called feature EB34 and costing $1,310 per machine, and the new drivers magically appear and you don’t have to attach your storage and other I/O peripherals to the Virtual I/O Server, which gives you a performance hit.

This is just speculation, of course. IBM is not going to help anyone figure out where potential issues could crop up. But I will say that generally speaking, I expect for IBM to start charging for the accelerated features on its Power and System z mainframes and that might mean deactivating them in its operating systems and hypervisors and only letting customers get access to them through software. This is exactly the kind of behavior you would expect from a Software Group that has taken over Systems and Technology Group, and a hardware vendor that needs to milk every last dollar out of every last transistor. By the way, I don’t expect any other chip makers that own their own software stacks–at this point, that is pretty much down to Oracle–to behave any differently. This is just business.

You can see the new License Agreement for Machine code here and see which makes have the “accepted by use” contract and which ones do not there.

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search