Reader Feedback On The Possibilities With IBM i Entry Systems Sporting Power8
September 30, 2013 Timothy Prickett Morgan
Great article Tim, you are definitely on the right track . . . but, I believe, you need to take it one step further.
We need to develop the successor to the IBM i operating system. (No, Linux is not the answer.) While incorporating all the truly wonderful aspects of i, (single level storage, SLIC, etc.), it needs to be designed from the ground up as a 2D and 3D graphical system. Likewise, we need a new event-based, object-oriented programming language. (No, not Java). ASNA‘s Visual RPG would make a fine starting point. Then, if done right, instead of having a system that would just compete with Wintel database servers, we’d have a truly superior system that could cover the gamut from tablets to enterprise business systems.
As I’ve elaborated on in multiple AS/400 Professionals Group LinkedIn posts, I firmly believe this could be accomplished with an IBM-funded skunkworks team that was independent of the IBM bureaucracy and politics. I also firmly believe it could be done at a funding rate of less than 1 percent of IBMs net profit per year. The payoff would be huge.
It is hard to not want to root for IBM to do some ground-breaking, forward-thinking systems redesign. Like you, I would love to see such a thing happen, to help redefine operating systems and the underlying hardware systems that run them. I just think a lot of these ideas are not coming out of the traditional systems vendors these days, but rather hyperscale data center operators, supercomputer centers, and academia. With IBM in particular, the company is so busy protecting its legacy platforms and services businesses and trying to build up an application software business outside of core ERP systems (marketing, content management, big data, and such) that it doesn’t want to do any redefinition of systems. IBM is all about preserving the customers it has with the systems they have and giving them just enough to keep them spending. IBM is not feeling radical, as it did in 1964 when it bet the company on the System/360 mainframe. Maybe you only get to do that once in a corporate lifetime. . . .
That said, I think IBM has plenty of systems expertise, and it could time the development of such a system as you talk about above with some major transition in technology. Memory and flash are about to run out of gas in two generations, disk drives are stalled in terms of performance, processors are stuck at their current clock speeds and just keep getting more cores and cache memory to use up the transistor budgets that Moore’s Law allows. I think you are right. Someone is going to have to do something to make a different kind of system with a different kind of software stack. I just don’t know if IBM will be that someone this time around.
I’d like to see another SAP workload study done with these new processors. I haven’t seen one in a while and it would be interesting to see where that compare to say a z/OS running on four IFLs, two ZIP/ZAAPs and the DB2 instances running on a general processors.
A price comparison would be interesting considering in my shop, we’re running all the guest on z/Linux and the databases on DB2. As I understand it, you can license OS/400 with/without DB2, so that might make an “all OS/400” implementation even cheaper.
I struggle to find SAP/CPW. Any conversion with existing or new processors?
It has been a long time since I saw any SAP Sales & Distribution benchmark tests from IBM on the OS/400-IBM i platform, but what I have been told by IBM’s performance experts who do the Business Warehouse benchmark tests is that the SAP SD test is largely a function of the application server performance, not the database performance. And because SAP runs in the PASE AIX runtime environment bundled in IBM i, for a given Power Systems machine, the IBM i performance will be “roughly comparable” to the AIX performance.
I think it would be interesting to show what that might be and how it compares to CPW and rPerf. So I will consider this a homework assignment.