> ... double ... to maybe 70?
Hah!
This article reminds me of new and amazing tech from years ago.
IBM introduced a specialty engine to help with particular workloads.
(Java specifically.) Knowing that the hardware had been micro-coded for
several decades AND seeing other specialty engines (like the Crypto
Express), I was excited about this new engine that could _interpret Java
byte code directly_.
Uhhh... no.
Turned out ZAAPs and ZIIPs (not to mention IFLs) were/are just
ham-strung general purpose engines.
So now we read that IBM will release "something" that will interpret ARM
instructions as well as Z instructions. Color me skeptical. (Something
about "fool me once", and such.)
Never the less, there it is in black-and-white.
Consider some facts:
* ARM has been making headway into the data center for several years
already
* IBM is struggling to sell Z instruction set and architecture,
especially for Linux workloads
* IBM is downplaying, abandoning, deprecating all Z environments aside
from z/OS
* Z instruction set is CISC, ARM is RISC
* they're talking about KVM with no mention of z/VM
Bullet #1, PHSiii may remember (and I certainly do) "the machine" from
when we were with HPE, a wholly ARM based multi-processor frame.
I have no idea where this project went after HPE spit us out (and we
became part of Micro Focus). I thought it was a slick idea, already
using ARM on my own LAN.
Maybe IBM manglement have seen the same ARM trend.
Bullet #2, sadly, the public doesn't get it. ARM competes with Z on
power efficiency, but does not have channelized I/O.
The public has no clue where that matters. And if you throw enough
resources at a problem, actual details matter less and less.
Bullet #3, consider geese and golden eggs, think "baby and bathwater",
which we on the outside (from IBM) grok intuitively.
Rewind to S/360, the first general purpose architecture but consistent
across the portfolio. But the only truly entrenched Z system is z/OS.
Bullet #4, maybe oil-and-water, or maybe complementary ... I really
don't know ... but they're seriously different.
Bullet #5, I use KVM *heavily* for my Intel/AMD farm. Works great.
Efficient? Dunno. Don't really care. (Because the host processor has
more power than I can use anyway.)
But fond as I am of KVM, it simply does not compare with z/VM. (The
latter being one of those things mentioned in bullet #3.)
If you're going to support context switching at all, you really want z/VM.
So what exactly are they going to do?
Give us a daughter board with a stack of ARM cores to sit alongside
Crypto Express?
That would work.
For that matter, why not give us an AMD/Intel daughter board and run MS
Windows workloads?
-- R; <><
On 8/24/26 2:09 PM, Phil Smith III wrote:
https://techstrongsemi.com/article/ibm-and-arm-unveil-dual-architecture-chip-plan-for-future-z-and-linuxone-mainframes
That might more than double the number of LinuxONE systems sold.to maybe 70?
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email [email protected] with the message: INFO LINUX-390 or visit
http://www2.marist.edu/htbin/wlvindex?LINUX-390
--
-- R; <><
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www2.marist.edu/htbin/wlvindex?LINUX-390
