Talk:Linux on System z

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Linux (Rated Start-class)
WikiProject icon This article is within the scope of WikiProject Linux, a collaborative effort to improve the coverage of Linux on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
Start-Class article Start  This article has been rated as Start-Class on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
 

31-bit or 64-bit[edit]

31 or 32 bits mode?

When Linux on zSeries is running on s/390 (31-bit hardware) then Linux can only run in 31-bit mode. When running on zSeries (64-bit hardware) then Linux can run in either 31-bit or 64-bit mode.
31. The mainframes never had a 32-bit mode. It is more accurate to say that the mainframes never had a 31-bit addressing mode (the registers were 32-bit however). It shifted from 24-bit to 31-bit (the last bit indicating the used addressing mode) with the ESA architecture.
Actually XA introduced 31-bit addressing and was supported by MVS v2 in the early 1980's. ESA introduced access registers increasing the addressing range horizontally rather than vertically. It was supported by MVS v3 in the later 1980s.
You're right for MVS/XA (I read my history notes). The first part of my post is accurate however. If I do not make mistake, access register are used to address address spaces?
Yes, everything you said was correct, esp if you substitute ESA with XA. access register mode allowed (and still does) addressing to address spaces and data spaces (which are like address spaces but were never originally an address space, so they only contain data as opposed to executable code, but can be shared or accessed by other address spaces ... esp if you know how to manipulate the address translation tables'-) Thus the `horizontal' analogy; being able to address multiple address/data spaces in 31 bit mode at the same time.
I apologize, these are my first entries to a wiki so I don't know how to attribute the comment to myself or how to format the discussion thread properly.
If you are Linas, then I had some correspondence with you while bigfoot was still going strong. Today I amuse myself working on the hercules emulator. --[gsmith]
I'm not Linas, sorry.
I wanted to refer to data space when I said address spaces... I never used those though. the only assembler programs i did were less than 4096 bytes :)
>I don't know how to attribute the comment to myself or how to format the discussion thread properly.
No problem, me neither.
About Hercules, I never tried to run the emulator. It seems complicated to set up. At my shop, we're installing some linux on Z/VM. With Hercules, I could theorically emulate a single S/390 Linux image legaly on my PC, no? --[fred]
IANAL, but I believe you are talking about running GPL'ed software, so I see no reason why not. --Maru (talk) Contribs 04:09, 5 November 2005 (UTC)
Thanks --142.213.160.67 14:36, 7 November 2005 (UTC)

NPOV?[edit]

This article reads very much like an IBM ad... --Liam Proven 19:54, 11 July 2005 (UTC)

Yes, shouldnt there be an "Criticism" paragraph? For instance, the zLinux has some problems: http://www.zdnet.com/blog/murphy/mainframes-ifl-and-linux/595
Probably because 80% of the anon contributors are IBM employees. But is there something in particular you object to? Something you think is incorrect? Don't just say its POV without saying what's wrong with it. --linas 22:20, 7 October 2005 (UTC)
Yes, this is pure marketing. I.e it has been stated the German Railway corporation is dumping it's Intel servers, in favour of going on zSeries. This is wrong. We bought hundrets of Intel and AMD64 Linux servers recently. The zLinux rollout is standing still, as it proved less powerful, problematic and instable. We have apart from some SAP appservers ONLY IBM stuff running on zLinux. The consolidation of Tivoli Storage Manager caused big downtimes. Lotus Notes was running on Windows before, not Linux. Now,after unsuccessfully deploying to zLinux we are moving the Notes servers from zLinux to Intel Linux. This articel is clearly representing the past (fantasizing and wish thinking of some people around our company), not the future. Cannot print my name here. So please do set an NPOV mark in the article !
Ok, I went through and improved it. Is it any more NPOV now? --Maru (talk) 01:57, 8 October 2005 (UTC)
While it seems slightly more neutral now, you've changed concrete statements to softer statements that aren't true. For example you said "IFL hardware tend to be less expensive than general purpose engines (CPs)."
There is no question that IFLs are always cheaper for running Linux than full CPs, hence the reason they exist. Saying they tend to be is incorrect.
You also added that "Like most other versions of Linux , Linux on zSeries is governed by the GPL"
As far as iI know, there are no versions of Linux that aren't governed by the GPL.
I have no idea what IFLs are compared to GPUs, so I erred on the side of NPOV. If you know what the deuce those are, then please feel free to change that. And there are indeed versions of Linux effectively not governed by the GPL. The GPL only applies to software publicly relelased (the P in the GPL). So Google's version of Linux could well be under some crazy Google license, since AFAIK they do not release the binaries publicly. If they did though, and tried to keep it under their crazy license, all hell would break loose, so it is a curious loophole, but largely of little relevance (though I hear GPL v3 is going to address services, closing that loophole). --Maru (talk) 04:29, 9 October 2005 (UTC)

That's not right, google is still obligated under the GPL. They cannot put their version of Linux under "some crazy Google license". (However, they do not have to disclose their source modifications). linas 04:47, 5 November 2005 (UTC)

Looks like the last little bit of edits put back a bunch of the marketing-speak. Jay Maynard 16:59, 12 May 2006 (UTC)

I disagree that this reads like marketing. IBM has a schizophrenic attitude toward zLinux. They want Linux users to migrate to mainframe Linux, but they do NOT want mainframe zOS or zVSE users to migrate to zLinux. Why? IFLs provide the same CPU capacity as general purpose CPUs at ~1% of the price, restricted to zLinux (and zVM). So, while this information is available, it is poorly understood, and this article goes a long way toward correcting that. I was very surprised at its thoroughness, and I think it is providing a unique and needed service. And I have absolutely no connection to IBM personally, though I do work with ports from mainframes to zLinux.

Okay -- sorry -- I have nothing against IBM, but Jay is right. This reads like a frigging brochure, not an encyclopedia entry. I think about 70 percent of this material should go. Rhombus 18:07, 20 January 2007 (UTC)

I'm not sure it should go, really...but it definitely needs a rewrite. If I get some free time this week, I'll take a crack at it. Jay Maynard 15:59, 21 January 2007 (UTC)

Hey, Jay, when and if you do...a lot of the language seems awfully familiar to me. I think pieces of it may have been cut-and-pasted from my z/Journal articles. I'll see if I can find specifics in the next few days. If this is in fact the case, well, there's your NPOV right there. My articles were never intended to be neutral: they're very much, "Hey, look at the neat things you can do with Linux on z" -- Adam Thornton

Here's an example: "System z earned EAL5+ certification, no other commercial system in the industry has been certified higher under the Common Criteria security evaluation." The first part of the sentence is fine. Everything after the comma is marketing. I'll be bold and remove it. Fordsfords (talk) 13:17, 6 May 2014 (UTC)

Running on Hercules[edit]

I think the existing language violates NPOV, since there's only one person who's ever raised the issue of whether Hercules violates any IBM patents. See, especially, the undue weight section of that page. IBM itself has never approached anyone with such a concern. Further, this page is not the place to discuss those concerns, as they're not specific to Linux.

I propose replacing the language with this:

The open source Hercules software is also available. While Hercules is not sanctioned for running licensed mainframe operating systems, such as z/OS, software licensing issues do not apply to running Linux on Hercules.

If someone thinks that the patent issues are that important, then an NPOV discussion can be added to the Hercules entry.

-- Jay Maynard 16:52, 15 May 2006 (UTC)

Section on emulators says FLEX-ES is "the officially sanctioned option" Says who? Shouldn't that be amended to say something more NPOV that it is a commercially supported option, rather than "official" in any respect? Further wrt to NPOV, I think it might be worth mentioning in the Price/Cost section that alternatives for virtualization exist on other platforms, so the consolidation benefits claimed for Linux on z are not necessarily unique to this one platform, and that the cost of porting an application from its original platform also need to be factored in. Jsavit (talk) 18:29, 30 December 2008 (UTC)

zUbuntu[edit]

The article should probably have something on zUbuntu, the Ubuntu port for zSeries. --Easyas12c 18:33, 15 June 2006 (UTC)

Is there anything beyond an original project announcement? Has any actual work been done? Jay Maynard 19:26, 15 June 2006 (UTC)
Well, there's a stub article at ZUbuntu :-) Seriously, it looks dead to me. RossPatterson 23:49, 17 August 2007 (UTC)

Large installation[edit]

While 1700 installs might be large considering how many mainframes that are still out there, it's not a large user base. —The preceding unsigned comment was added by 216.136.85.66 (talk) 01:19, 7 December 2006 (UTC).

Not a very large user base but I've taken the liberty of pointing out that each system might be supporting thousands of "seats". MarkMLl 23:42, 17 August 2007 (UTC)
Most of these systems are trial installations. Then they were counting zLinux Server instances not zSeries machines. So Formerly 1700 installations will be perhaps 170 Machines.

By now, there are less if you are honest. Unfortunately true numbers will not be published, neither by IBM, nor by those few customers who were so keen to try it out in production.

I've cut the quote down to a strict statement of the facts: that IBM claimed a certain number of installations at a certain time. This much can be supported by the cite given. JeffLicquia (talk) 20:10, 23 October 2013 (UTC)

History[edit]

Sorry but I've tagged part of this as dubious. Please could the original author justify the statement "This 100% open source status is unusual among Linux distributions, many of which still contain OCO drivers from various vendors who wish to hide proprietary implementation details from the Linux community". Granted that on many architectures there is a small number of peripheral controllers that require proprietary microcode but I don't see IBM GPLing channel controller programs/firmware/microcode so in practice there's very little difference. MarkMLl 23:54, 17 August 2007 (UTC)

I slightly revised this language. And no, we're not talking about microcode, BIOS, etc., here, which almost nobody publishes or provides as open source. The old and, better yet, new language makes it clear we're talking about drivers. And yes, this is a big deal to the Linux community, closed source drivers. So the new language has two changes: one is to repeat the word "drivers," since that's what's being hidden. The second is to remove the "from the Linux community" part since, in fact, closed source drivers hide implementation details from everyone. -- BBCWatcher —Preceding unsigned comment added by BBCWatcher (talkcontribs) 06:57, 12 September 2007 (UTC)

Article quality[edit]

Simply one of the worst Wiki articles I have read. No Encylopedia article should ever include slang or non-standard grammar conventions. This whole article is full of very isolated slang used in the "xxx" format and all of it should be removed. In fact all of the "xxx" points should be removed period as the convention voilates professional writting conventions unless completely necessary which is to say almost never needed. Also, from working in IT I get the impression the author does not actually know what he is talking about. Any article written for an encyclopedia should not convey any opinion or emotion at all - it should be just the facts. This article is clearly not that.

I strongly suggest a complete rewrite —Preceding unsigned comment added by 157.203.43.103 (talkcontribs) 28 November 2007

I guess it has been started by IBM marketing dept. Is it better now? I think a rewrite is an overkill for such low-traffic article.
Btw, it seems you understand wikipedia guideline to be bold to literally :)) --Kubanczyk (talk) 14:58, 28 November 2007 (UTC)

Performance[edit]

Not sure how or if to interject performance into this story. Linux on zSeries is both a nice bit of work, and a bit of a joke. The nice part comes of getting Linux to work on mainframe hardware. The joke comes from the extremely weak performance of zSeries CPUs in comparison to common desktop CPUs. What this means is that programs that run well on current mainstream desktop or server hardware - and use significant CPU - will perform very poorly on mainframe hardware. Attempting to use mainframe hardware to host web servers - for anything other than static files - is likely a mistake.

At least this was true when I had access to zSeries hardware. Is this still true?

IBM used to disallow publishing benchmark results. Do they still? If so that means finding references for the Wikipedia article is unlikely. pbannister (talk) 21:49, 26 May 2010 (UTC)

I'm not aware of any published benchmarks comparing zLinux to x86Linux, but I haven't looked either. As to your suggestion that non-static websites do poorly on the mainframe, is http://www.nationwide.com a static webpage? It's served up by a IBM z9 mainframe running Linux. --TreyGeek (talk) 22:15, 26 May 2010 (UTC)
The lack of public benchmarks is by no means proof, but is a pretty strong clue. When you have a good performance story, you want everyone to know. (Did poke around a bit, and came up essentially empty - though there are strong hints: mainframe vs Linux, licensing costs, refuting IBM exaggerations). As to the nationwide.com site, did they make the mistake of paying far to much for the capacity they need? pbannister (talk) 01:13, 2 June 2010 (UTC)
This is important to know for those considering to virtualize Linux on Mainframes. If a company think they can virtualize number crunching, they should know this weakness. I have created a header "Disadvantages" that tells this story. —Preceding unsigned comment added by 217.73.15.6 (talk) 08:21, 28 September 2010 (UTC)
It is also important to provide a source for information you add to Wikipedia. --TreyGeek (talk) 08:49, 28 September 2010 (UTC)
Yes, you are absolutely right. Here are three links. Here is a source from Microsoft http://www.microsoft.com/presspass/features/2003/sep03/09-15LinuxStudies.mspx? "we found that each [z9] mainframe CPU performed 14 percent less work than one [single core] 900 MHz Intel Xeon processor running Windows Server 2003." The z10 is 50% faster than z9, and the z196 is 50% faster than z10, which means a z196 is 1.5 x 1.5 = 2.25 times faster than a z9. This means a z196 corresponds to 2.25 x 900MHz = 2 GHz Intel Xeon. But todays modern server x86 cpus have 8 cores, which means they have in total 8 cores x 2 GHz = 16 GHz. We see that x86 at 16GHz is more than z196 at 2GHz. This shows that a z196 is not suited for scientific calculations nor number crunching. Here is another source from a famous Linux expert that ported Linux to IBM Mainframe, who says 1MIPS == 4MHz x86. http://www.mail-archive.com/linux-390@vm.marist.edu/msg18587.html This shows that a z196 with 1400 MIPS corresponds to 5,6GHz x86. But a modern x86 has 8 cores, that means it has in total 16GHz, which is 3x faster than 5.6GHz. Again, we see that the Mainframe is not suited for number crunching. Here is another link where the cofounder of TurboHercules says that a 8-way Nehalem-EX gives 3.200 MIPS using software emulation: http://en.wikipedia.org/wiki/TurboHercules#Performance But software emulation is 5-10x slower. This means a 8-way Nehalem-EX running native code should be 5-10x faster, that is, 16.000 - 32.000MIPS. This big MIPS number matches a fully equipped z196 mainframe with 24 cpus. Again, we see that the Mainframe is not suited for number crunching. Now, as you deleted my text because I provided no links, please add my text again, as here you have three links. The customers will be disappointed if they try to virtualize number crunching on Mainframes, they need to know this. Dont you agree? —Preceding unsigned comment added by 217.73.15.6 (talk) 11:34, 30 September 2010 (UTC)
Why is this statement here in "Appropriate workloads" it looks weird: "mainframes in general, and Linux on zSeries in particular, do not perform well, at least on a cost basis, for single task computations"? The three links above shows that Mainframes do not perform well for single task computations. It is not about "at least on a cost basis" - it is about Mainframes do not perform well at all for number crunching. Please remove "at least on a cost basis" which might fool the reader that if it was free to compute anything on a Mainframe, the Mainframes would excel on number crunching. But they dont. Read the three links above. —Preceding unsigned comment added by 217.73.15.6 (talk) 13:07, 18 November 2010 (UTC)

Why was this part of assessment of emulation performance removed? It was very relevant for people wanting to know more about what emulation can give. Was that part removed by an IBM salesperson wanting to hide this information? — Preceding unsigned comment added by 213.114.157.231 (talk) 18:32, 27 December 2013 (UTC)

Free software cite[edit]

I'm at a loss as to what would constitute a cite for the following statement: "However, currently Linux on System z is completely free software under the GNU General Public License." (currently marked as "citation needed") The s390 and s390x architectures are in the mainline kernel, and are used to build kernel packages suitable for Debian main, Fedora, etc. Ditto for the support packages (libc, gcc, binutils, etc.). Is it good enough to just link to, say, the s390 tree on kernel.org? JeffLicquia (talk) 20:21, 23 October 2013 (UTC)

OK, since no one seems to know, I'm removing "citation needed". If you disagree, please discuss what a cite would look like. JeffLicquia (talk) 18:48, 29 August 2014 (UTC)

Virtualization[edit]

I propose removing all paragraphs in this section except the first, second, and sixth, and heavily editing the sixth paragraph. There are too many uncited technical details. I'm pretty sure I can find cites for the first paragraph, but if not, then maybe that one needs to go as well. JeffLicquia (talk) 19:35, 29 August 2014 (UTC)