On 05/01/2018 06:06 PM, [email protected] wrote:
On Tuesday, 1 May 2018 07:38:52 PDT Roland Hughes wrote:
For decades the x86 defined the absolute bottom of the IT universe. When
you didn't care about it you put it on x86.
You do realise that 99% of the Cloud today is x86, right? Of the rest of the
1%, I'm going to guess it's 50-50 between ARM and IBM POWER, with only the
latter usually running in big-endian mode. So right about 99.5% of the cloud
and Internet datacentres are little-endian.
LOL,
And I still hear Chicago Bears fans singing the "Super Bowl Shuffle,"
but that glory day passed and will never return again.
https://www.pcworld.com/article/3178387/servers/microsoft-ports-windows-server-os-to-arm-with-help-from-qualcomm.html
Okay, admittedly, MS had yet another EPIC fail there. ZUNE-3 as it were
with ZUNE-2 being Windows Mobile and Microsoft Bob is in a distant
category keeping Microsoft Money company. Both of them are constantly
bothered by this gnat called Clippy. If those names mean nothing to you,
consider yourself a child and go look them up. The failure here was pure
Microsoft, not the Qualcomm side of life and the rest of the industry
saw that.
http://www.datacenterknowledge.com/hardware/red-hat-bets-data-centers-are-ready-arm-servers
https://www.ubuntu.com/server Runs on all major architectures – x86,
x86-64, ARM v7, ARM64, POWER8 and IBM s390x (LinuxONE)
*Ah LinuxONE. Where SuSE and Ubuntu go to run clouds.**A big-endian z-box.*
https://www.zdnet.com/article/the-mainframe-lives-on-in-ibms-linuxone/
https://www.ibm.com/blogs/systems/the-next-phase-of-secured-computing-ibm-cloud-private-on-z/
https://www.wired.com/insights/2014/04/back-future-cloud-wont-replace-mainframe/
https://www.computerworld.com/article/2872096/ibm-s-z13-and-the-case-for-the-mainframe-cloud.html
https://www.wired.com/2015/01/z13-mainframe/
https://www.suse.com/c/mainframe-versus-server-farm-comparison/
http://enterprisesystemsmedia.com/article/mainframes-are-still-at-the-heart-of-the-modern-tech-world#sr=g&m=o&cp=or&ct=-tmc&st=(opu%20qspwjefe)&ts=1525461227
Data centers have been moving to ARM (the new absolute bottom) and
Z-machines for quite some time now. The x86 has both heat and power
problems which cannot be overcome. Even HP switched to selling ARM based
servers in 2014.
http://www8.hp.com/us/en/hp-news/press-release.html?id=1800094#.Wu20CtYh2V4
Though IBM has been pushing Power into the low end market and there is a
legitimate reason for the low end market to consider that path. Once
your server farm exceeds size X you can transition to a Z-box without a
lot of pain reducing both power consumption and cooling requirements of
the data center. With x86 and ARM one is screwed. You have to build a
different data center connected to a different power grid. There is no
path up. Well, the x86 based server farms first migrate to ARM,
drastically reducing power and heat, but, not permanently moving the
goal post.
Even Google dumped INTEL.
https://www.bloomberg.com/news/articles/2016-04-06/google-taps-ibm-rackspace-to-dent-intel-s-hold-on-server-chips
Google voice has been ported to a Z-machine. It was squirreled away in
one of the above links.
Da Bears - day ain't in da Super Bowl.
Da Fridge - don't play football no more.
Da Coach - don't live in da city no more.
Da x86 - just a shadow in Da Datacenter now.
Eventually Bears fans will realize it's not 1986 anymore. Eventually x86
fans will realize it's not 1999 anymore. Until then we will have these
discussions.
While all of this is an entertaining conversation, it doesn't change the
fact that big-endian ___must___ remain the default binary format for the
class in question. That class was initially created for the sole purpose
of communicating back to real computers. Your personal view of the chip
universe (or mine for that matter) does ___NOT___ change the purpose of
that class. I don't even like IBM! I grew up on DEC hardware with OpenVMS.
http://www.theminimumyouneedtoknow.com/app_book.html
https://en.wikipedia.org/wiki/OpenVMS
It's up or out in the IT world. For decades the x86 defined the absolute
bottom of the IT industry. It grew consuming its betters which refused
to move up. Remember Wang? The mini-computer giant whose market segment
was collaborative word processing. During the late 1980s the PC got
WordPerfect and people lived without real-time collaboration. Wang went
bye-bye. What's the biggest selling point for word processing today?
Real-time collaboration. The need didn't go away, just the willingness
to pay for it.
Singer used to make computers.
https://en.wikipedia.org/wiki/Singer_System_10
Then along came DEC with the PDP 11 and RSTS/E with VT52 terminals,
quickly followed by real VT-100 terminals, not the absolutely worthless
"vt-100" emulators in the Linux world, but actual VT-100 terminals which
worked with both RSTS/E and VMS. The software things in the Linux world
instantly get someone designated as a "maintainer" who disables roughly
half of the keyboard/emulator functionality because it conflicts with VIM.
At any rate, the ERP package (Perspective 500, 800, etc.) I worked on
came from a Singer BASIC package. That's how I know that Singer, famous
for sewing machines, also used to make business computers.
I'm old, so I could go on about the various niches midrange and
mainframe computers had and why they failed. Hell, DEC was the first and
only midrange which could talk with every computer back in the day. They
spent a lot of time and money designing DECnet, making it secure and
reliable. It wasn't free though. Eventually customers decided they would
just use this new fangled TCP/IP stuff on a near "free" OS allowing
hackers from around the world to steal customer data, identities,
engineering specs, whatever they wanted, because it was cheaper than
paying for something. (After Facebook, that pendulum is going to swing
the other direction for a while though.)
The x86 is sitting ___EXACTLY___ where Wang was.
The bottom end can do waaaay more for a fraction of the
cost/electricity/heat. The x86 can't even dream about chasing the
transactional I/O throughput of a Z-box, in large part, because that
requires building an entire machine, not just a chip. Lots of the market
has moved to tablets which are already ARM. Corporate desktops are
moving (have moved) to notebooks/laptops, many of which are now using ARM.
It's odd. OpenVMS found its way into most IBM owned data centers because
engineering teams needed a feature only OpenVMS had, like the VMSDate
timestamp used for NASA stellar drift calculations. Today the last
clinging vestige of the x86 in corporate America is the engineering
department. Why? Many of us doing embedded work still need an __actual__
serial port. Some equipment requires an __actual__ parallel port.
Thankfully HP still has a few engineers specking out their hardware.
HP EliteDesk 705 Microtower with serial and parallel ports.
https://store.hp.com/us/en/mdp/desktops/elitedesk-700-microtower-244031--1#!&tab=features
705 small form factor
https://store.hp.com/us/en/mdp/desktops/elitedesk-700-small-form-factor-244032--1#!&tab=features
The downside for the x86 is that most of us are also starting to need
CUDA for extreme localized crunching. At some point these machines will
also switch to ARM and CUDA. When it comes to off-chip number crunching,
CUDA won the war.
--
Roland Hughes, President
Logikal Solutions
(630)-205-1593
http://www.theminimumyouneedtoknow.com
http://www.infiniteexposure.net
http://www.johnsmith-book.com
http://www.logikalblog.com
http://www.interestingauthors.com/blog
http://lesedi.us/
http://onedollarcontentstore.com
_______________________________________________
Interest mailing list
[email protected]
http://lists.qt-project.org/mailman/listinfo/interest
