Andrew Rowley wrote: >IBM were saying how great it was at parallelizing work, however that >is a bit of a red flag for me.
OK, but one must mark such instincts to market. IBM introduced zAAPs just over 13 years ago, and one site quickly put that technology into production on September 1, 2004. In their case they doubled their execution threads (as I recall), literally overnight, in the late night hours across August 31, 2004, and September 1, 2004. Later, zIIPs supplanted zAAPs, and starting with the IBM z13 each zIIP now has two threads of execution (SMT2). If you run non-trivial Java workloads on z/OS, you're virtually guaranteed to have at least three physical threads of execution for those workloads on IBM z13 and higher machine models. If a workload can take maximum advantage of all these CP+zIIP threads, terrific. Also, mainframers (and others) frequently use the word "parallel" colloquially to refer to job steps, particularly in batch, that can be scheduled to run at the same time -- that step #2 doesn't have to wait for step #1 to complete its work. So if step #1 is waiting for some I/O or whatever, no problem, even a single processor can give some attention to step #2. And that's also terrific. If you're trying to optimize program execution, "parallelizing" each step insofar as possible is magical stuff. As another example, IBM's Scalable Architecture for Financial Reporting (SAFR) uses such language, colloquially at least. Heck, even "uniprocessor" System/360 and especially System/370 machines embodied parallel execution principles, especially in the I/O subsystems and channel architecture. It's terrific that the I/O devices can and do carry out their responsibilities, in parallel, while the main processors do something else. Anyway, I get your point, that the P word might have caused some anxiety in the past. It the past (and to some diminishing extent now) it meant throwing more CPU at a problem to solve it. When one is focused on resource efficiency and scalability then throwing any resources at a problem is not ideal. But I wouldn't over-interpret that word, especially nowadays. Thank goodness work can be parallelized. -------------------------------------------------------------------------------------------------------- Timothy Sipples IT Architect Executive, Industry Solutions, IBM Z and LinuxONE, AP/GCG/MEA E-Mail: sipp...@sg.ibm.com ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
