Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
Hello On Tue, Mar 25, 2008 at 03:12:13AM +0100, Enrico Weigelt wrote: * defect management directly on block basis (w/o additional stacking layers). IMO dividing things into layers/parts is good. It allows for replacing one layer, or not using some of them if they are not needed. * distributed storages (not just disks, but several hosts) That sounds interesting. * volumes attributes which let the volume manager decide how/where to actually store blocks, eg.: Is there a real need for this? (Just wondering) -- Wait few minutes before opening this email. The temperature difference could lead to vapour condensation. Michal 'vorner' Vaner pgpBQEUlYqGXQ.pgp Description: PGP signature
[gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
Hi folks, after reading several articles about Mainframes and similar archs (even ancient ones like B7000), I wonder if Linux world could learn something from there. One very interesting point (IMHO) is the storage abstraction. AFAIK, Mainframes work on one large virtual memory (disks for swapping out RAM, tapes for swapping out disks, etc). This way you just allocate some piece of space (like some virtual partition) to an application (of guest). If you need more space, just plug in more disks and the OS will handle all this automatically. I'm currently planning to implement an similar approach for Linux (at least virtual block devices). What do you think about this ? cu -- - Enrico Weigelt== metux IT service - http://www.metux.de/ - Please visit the OpenSource QM Taskforce: http://wiki.metux.de/public/OpenSource_QM_Taskforce Patches / Fixes for a lot dozens of packages in dozens of versions: http://patches.metux.de/ - -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
Enrico Weigelt wrote: Hi folks, after reading several articles about Mainframes and similar archs (even ancient ones like B7000), I wonder if Linux world could learn something from there. One very interesting point (IMHO) is the storage abstraction. AFAIK, Mainframes work on one large virtual memory (disks for swapping out RAM, tapes for swapping out disks, etc). This way you just allocate some piece of space (like some virtual partition) to an application (of guest). If you need more space, just plug in more disks and the OS will handle all this automatically. I'm currently planning to implement an similar approach for Linux (at least virtual block devices). What do you think about this ? cu Check out LVM (Logical Volume Manager) http://tldp.org/HOWTO/LVM-HOWTO/ http://www.gentoo.org/doc/en/lvm2.xml Seems to do exactly what you're talking about. Chris -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
On Mon, 2008-03-24 at 16:03 +0100, Enrico Weigelt wrote: Hi folks, after reading several articles about Mainframes and similar archs (even ancient ones like B7000), I wonder if Linux world could learn something from there. One very interesting point (IMHO) is the storage abstraction. AFAIK, Mainframes work on one large virtual memory (disks for swapping out RAM, tapes for swapping out disks, etc). This way you just allocate some piece of space (like some virtual partition) to an application (of guest). If you need more space, just plug in more disks and the OS will handle all this automatically. I'm currently planning to implement an similar approach for Linux (at least virtual block devices). What do you think about this ? I am not certain this is the true for mainframes, at least not all of them. But interestingly enough the Unununium project had a similar philosophy, basically L1/2 cache, RAM, and disk were essentially the same things, though with different price/performance ratios, and that each should be indistinguishable for the user. Personally I don't think that level of abstraction provides any great benefit for the user, though from a strictly technical standpoint it is at least interesting. If you are speaking strictly of hot-pluggable memory/storage then Linux has this already (if the hardware supports it), and at least Xen gives a similar mainframe type feeling for allocating/deallocating storage/memory for guests on-the-fly. -a -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
On Monday 24 March 2008, Enrico Weigelt wrote: Hi folks, after reading several articles about Mainframes and similar archs (even ancient ones like B7000), I wonder if Linux world could learn something from there. One very interesting point (IMHO) is the storage abstraction. AFAIK, Mainframes work on one large virtual memory (disks for swapping out RAM, tapes for swapping out disks, etc). This way you just allocate some piece of space (like some virtual partition) to an application (of guest). If you need more space, just plug in more disks and the OS will handle all this automatically. I'm currently planning to implement an similar approach for Linux (at least virtual block devices). What do you think about this ? I believe these concepts are to a large degree implemented in an existing product. It's called ZFS. Pity about the license though. We are out in the cold. -- Alan McKinnon alan dot mckinnon at gmail dot com -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
Alan McKinnon wrote: On Monday 24 March 2008, Enrico Weigelt wrote: One very interesting point (IMHO) is the storage abstraction. [snip] I'm currently planning to implement an similar approach for Linux (at least virtual block devices). You might want to check out Plan 9 from Bell Labs: http://plan9.bell-labs.com/plan9/ Have fun, Roy -- gentoo-user@lists.gentoo.org mailing list
Re: [gentoo-user] 1/2 OT: What Linux could learn from mainframes ?
* Chris Frederick [EMAIL PROTECTED] wrote: Check out LVM (Logical Volume Manager) http://tldp.org/HOWTO/LVM-HOWTO/ http://www.gentoo.org/doc/en/lvm2.xml Yes, at least for the storage stuff, LVM2 can do much of this. But my ideas go some steps futher, eg: * mapping blocks instead of larger chunks * defect management directly on block basis (w/o additional stacking layers). * distributed storages (not just disks, but several hosts) * volumes attributes which let the volume manager decide how/where to actually store blocks, eg.: + mirroring: min/max number of copies, allowed device classes + compression: allowed algos or grades + encryption: ciphers, keys, ... + reliability: hard/limit (eg. use unstable disks exceptionally) + caching: allow some volumes to be cached on fast devices ... cu -- - Enrico Weigelt== metux IT service - http://www.metux.de/ - Please visit the OpenSource QM Taskforce: http://wiki.metux.de/public/OpenSource_QM_Taskforce Patches / Fixes for a lot dozens of packages in dozens of versions: http://patches.metux.de/ - -- gentoo-user@lists.gentoo.org mailing list