Submitted by Sources close to Canonical and IBM are telling me that Ubuntu may soon be certified on IBM’s System p mini-computers and blades and System z mainframes.
Sources close to Canonical and IBM are telling me that Ubuntu may soon be certified on IBM’s System p mini-computers and blades and System z mainframes.
Let’s dist-upgrade the mainframe every 6 months or even better update-manager-gnome -d
Or use the long term support versions.
I highly doubt that this is certifying Ubuntu to run a baremetal install. Instead, this will be running Ubuntu inside the mainframe equivalent of a VM (or LPARS or whatever the IBM terminology is).
This is really no different than Ubuntu being certified to run in MS Hyper-V. It’s just a guest OS install.
Since Linux can run baremetal on a System z, it will be quite strange that this certification only cover the z/VM.
I could be wrong, but RedHat and Novell looks to be certified to run as a baremetal install, thus Ubuntu will not be the first case.
RedHat and Suse don’t run baremetal either.
But on System Z, as I understand it, most operating systems (must ?) use paravirtualization.
Which means special drivers for the special virtual devices which are available on system Z.
Linux can run on the mainframe, natively, baremetal.
In the article, one comment says “IBM provides IFL which is a chip specifically for running Linux on the mainframe with or without zVM”. Then we can see it graphically in
http://www-03.ibm.com/systems/resources/systems_z_os_linux_solution…
That image is from the page http://www-03.ibm.com/systems/z/os/linux/solutions/ifl.html, where you can search for the text “native Linux”.
Ohh, I didn’t know that, seems they do that since 2000.
A special chip ?
I guess that means a special CPU-microcode ?
Yep, just checked on Wikipedia it says:
“The specialty processors are all identical and IBM locks out certain functions based on what the processor is characterized as” … “reduced cost”
Seems the special chips are optional, has stuff disabled so you can only use it for Linux but at a reduced cost.
I read from several places that all IBM Mainframe OSes are running ontop a hypervisor, that no OS runs baremetal.
Are you sure that Linux can run baremetal on IBM Mainframes? Do you have a reference or link?
In http://www-03.ibm.com/systems/resources/systems_z_os_linux_solution… we can see that in those mainframes you can run Linux inside a “virtual machine” (using z/VM) … or you can run Linux without a virtual machine (as a native operating system).
If you visit, with Javascript enabled, this page:
– http://www-03.ibm.com/systems/z/os/linux/solutions/ifl.html
– you click on the “No additional IBM software charges for traditional environment” link,
– you search for “native Linux”,
– you see the mentioned diagram afterwards.
In my case, I visited, with Javascript disabled, this page:
– http://www-03.ibm.com/systems/z/os/linux/solutions/ifl.html
– searched for “native Linux”,
– saw the mentioned diagram afterwards.
I guess it is Ubuntu Server which they want to certificate?.. and yes, it possible to certify it to run on a LPAR.
Or possible, if they are looking to run Ubuntu Workstation, maybe is for a Remote Workstation solution with IBM z servers.
Oh how very interesting! my my, indeed!
(just being a troll haha)
So does this mean the PPC builds will be upgraded to fully official status again? And System Z? I didn’t know Ubuntu made s390 builds at all…
Mainframe cpus are much much slower than x86 cpus. Any decent Quad core x86 cpu is much faster than the newest and fastest Mainframe cpu. So why would anyone want to run Linux on slow extremely expensive hardware?
Look at dhrystone mips, and compare the latest and fastest z196 Mainframe cpu vs Quad Core i7: 50.000 MIPS vs 160.000MIPS
http://en.wikipedia.org/wiki/Instructions_per_second#Timeline_of_in…
I think because a System Z10 is fully redundant (including for example memory) and can run many, many Linux machines.
Sometimes you don’t need the all that CPU-power just a lot of different machines.
Running just the ‘one’ system is a lot more power and cooling efficient than a large number of Intel-based machines.
Some even support water-cooled.
And because of the redundancy it is very reliable.
Those aren’t drystone MIPS in the table for z196. They are the zMIPS mentioned earlier in the article. You are comparing different units and drawing incorrect conclusions.
Yes I know it is different MIPS, I am aware of that.
However, there are other links that show IBM Mainframe cpus are slow, as well.
Here are some numbers that show how slow IBM Mainframe cpus are:
http://www.osnews.com/thread?419352
People who keep coming into these topics and talking about just cpu power seriously need to do some research.
Because in the enterprise sector raw cpu power doesn’t mean nearly as much as i/o ability and redundancy. Most servers actually sit with the cpu idling while waiting for disk or network access. This is why virtualization is becoming so common.
Secondly, mainframe architecture is very different from small systems. In a small system the cpu handles darn near everything, so it needs to be high powered. In a mainframe, all i/o is relegated to specialized processors, and there are specialized processors for Java and database programs.
A mainframe is in fact a cluster, so tightly integrated that only the head sysadmin needs to know/care.
So what? I claim that IBM Mainframe cpus are several times slower than a high end x86 cpu. I dont see that you deny that. You are talking about good I/O, etc – I have never talked about Mainframe I/O. Are you trying to shift focus away from the slow Mainframe cpus, and try to make me say things I never said about I/O?
Thus, the IBM Mainframe cpus are several times slower than a high end x86 cpu. Look in a post higher up, by me, for more information about the slowness of IBM Mainframe cpus.
Here is more information about the latest IBM Mainframe cpu, the Z196, which runs at 5.26GHz and has close to half a GB of cpu cache. IBM dubs this the “Worlds Fastest CPU” – which is quite dubios advertising.
http://www-03.ibm.com/press/us/en/pressrelease/32414.wss
PS. I dont understand how IBM can claim that a big Mainframe can virtualize 1.500 x86 servers, when there is no way that 24 slow Mainframe cpus can do the work of much faster 1.500 x86 cpus?
So, what, is it your goal to bring this textbook strawman argument of yours into every news which touches on mainframes? Glance over replies every time?
Despite it being pointed out at least once ( http://www.osnews.com/comments/25046 and I don’t see why to “agree” with something which isn’t ever your initial premise, in fact is a rephrasing of what needs to be largely pointed out every time) to you what kind of fallacious point it is and that nobody promotes mainframes as if they were about CPU power, nobody uses mainframes like that, that’s not their point …yet you must criticise them on this non-issue. Nobody seems to treat them as such except you…
You are being dishonest in your evasions. Your premise does boil down to “why would anybody care about mainframes at all, their CPUs are so slow!” …yes, it presents things like only CPU would matter, like you would dismiss all the other factors.
It’s like (inevitable car analogy :p ) you’d bust into some article about a nice articulated, low-floor hybrid bus, and disparage it on the basis that it is slower than a racecar; building on that basis alone your wonder why anybody would choose to ride on it, why public authorities would waste money on buses for public transport instead of simply buying racecars.
(and many servers, services, are idling vast majority of the time, waiting for “random” request which need to be speedily replied to; that’s probably a good case for “excessive” virtualisation)
Edited 2011-09-14 23:57 UTC