Re: Links to decent 'why the mainframe thrives' article

On Jul 14, 2007, at 9:01 AM, John S. Giltner, Jr. wrote:
------------------SNIP-------------------

The "Mhz" rating of a zSeries CPU is slow than most of modern processors. However that does not mean a computer using a slower processor is slower than a computer using a faster process. That is like saying just be cause the engine in one car runs at a higher RPM makes it faster than a car whose engine that runs at a lower RPM. The problem is that ignores the transmission and differential gearing.

In this case it ignores the CPU's various instruction sets. Say you have two CPU that have different instruction sets. During a test it was found that CPU#1 can execute a mixed set of it instruction at the rate of 100 MIPS, and CPU#2 tested at 200 MIPS. Which is faster? You can't tell. The problem is that CPU#2 may have a less efficient set of instructions and you may have to execute 3 instructions to do the same amount of work as CPU#1 does with 1 instruction.

I remember reading a article about the FLEX/ES software when it first came out. They were taking about the difference in the two instruction sets. The lowest instruction ratio was 1:2, zSeries to Intel, the maximum was 1:1200 and the average was 1:17. So on average for every 1 zSeries instruction they had to execute 17 Intel instructions to do the same work.

You also have the fact that some functions on zSeries are handled by the hardware. Memory protection for one, zSeries hardware prevents one task from getting to another tasks memory. I know on the Intel platform this must be handled by the OS, so this is more instructions that the OS must execute and more work done by the CPU.

IBM published a paper on the differences between zSeries and PowerPC and the fact that zSeries CPU's are much better for SMP environments that run varied workloads on a single computer. This deals with the way that L2 cache works and what happens when you have a context switch. On the distributed platforms L2 is not shared (pre-multi-core).

Some distributed OS's will actually hold up work if the CPU it was last dispatched on is not available right now, this is to prevent the context switch on another CPU.

John,

Any chance this is online & where?

Ed

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@xxxxxxxxxxx with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
.

Relevant Pages

  • Re: Links to decent why the mainframe thrives article
    ... those used in other platforms. ... In this case it ignores the CPU's various instruction sets. ... So on average for every 1 zSeries instruction they had to execute 17 Intel instructions to do the same work. ...
    (bit.listserv.ibm-main)
  • Re: [PATCH] x86 - Enhance DEBUG_RODATA support - alternatives
    ... has been pulled out of the x86 tree. ... text_poke required to support this. ... correctly and so the CPU HOTPLUG special case can be removed. ... When you use this code to patch more than one byte of an instruction ...
    (Linux-Kernel)
  • Re: Simple function arguments
    ... are 2 names refering to the same memory location and use that. ... In the internals of a CPU there are various registers. ... address is stored from where the next instruction from memory is read and executed. ... what is generally referred to as 'The stack'. ...
    (comp.lang.cpp)
  • Re: wikipedia article
    ... parallel but skewed by one instruction. ... If the first CPU instruction execution causes a miss, ... memory access. ... distinguish between instruction and data references, ...
    (freebsd-questions)
  • Re: How does this make you feel?
    ... >>>primitives to implement, say, a memcpy just as efficiently as microcode ... > The work is offloaded from the programmer in any case - this type of code ... library macros need updating for new CPU products, ... And designing such instruction such that they don't ...
    (comp.arch)