Heterogenous, dynamic RAID for a home server
Hey people, A friend of mine want to build a file server machine at home to get the benefits of a Drobo, but for 1500 NIS rather than $1500. The idea is to mix in 1T, 2T and 3T disks he already has, and on occasion grow the ׁ•torage by another disk or replace and rebalance it. He wants it to be more reliable than RAID0 but more efficient than RAID10, so RAID 5/6 are more like it. I looked around at the idea of ZFS or a distributed FS on a single node, but they are not options. I'm thinking how to do it with a block redundancy scheme, and so far I came up with this: create a RAID5 from the lower 1TB of all the disks you have, then a smaller RAID5 array from the second TB, and finally a RAID1 or RAID5 from whatever you have left. so if we say I have 2*3T and 2*2T and 1*1T, I have (5-1)*1T+(4-1)*1T+1T (the last is RAID1). now I have 8TB in three block devices, and I make a single LV (LVM2) from three extents. Is there a better solution? do note that it's not optimal for spindle activity, but this is a home machine serving two other computers and a TV streamer at the worst case. Any feedback is welcome... -- Perishable item Ira Abramov http://ira.abramov.org/email/ ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
Re: Preparing to convince to shift to non-propriety documents formats
On Wed, Feb 22, 2012 at 00:45, Oron Peled o...@actcom.co.il wrote: Regretfully, your noble suggestion does not give any significant protection, for various reason (IANAL): * Patents: control *use* and not implementation. So if you write patent infringing code, you have no problem as long as you don't run it. However, your users are at a risk. (as a demo, see how MS threaten OEM's for Android use and not Google) * You may think about idemnifying your users (i.e: promise to financially back their damages), but this isn't a reasonable option unless you have spare cache in the 5-6 digits range (USA, in dollars). * Copyrights: this is a lesser risk, since we know free software developers do not copy/derive code from MS. However, even in this case -- if you are sued for copyright infringment, there's nothing that protect your users from being sued also (In the USA the MPAA/RIAA reminds everybody of this fact -- they sue the end users even if they downloaded infringing content from other infringing party, like youtube/pirate-bay/etc.) Thank you for your insight. I had figured that it is the distribution making available the code which could be construed as infringing. That means that the code will be released under GLP but the copyright remains with me, not you. But I think you guys know me, my intention is only to protect the real author, not to profit from the code. Profit from free software is not a shame. On the contrary, the client gets the program he asked for and as a bonus it's free software -- So the client gets better value for money. That's why I'm really sorry to hit your inovative bussiness model. I wish it would be feasible. No business model, I had not intended on making any money. Quite the opposite, I had intending on taking a risk with no financial incentive. ...The true author can remain anonymous if he wishes. Again, this isn't too practical these days (with BB anywhere). BB? -- Dotan Cohen http://gibberish.co.il http://what-is-what.com ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
Re: Preparing to convince to shift to non-propriety documents formats
Again, this isn't too practical these days (with BB anywhere). BB? Sorry, I did not have my 1984 hat on. Big brother. -- Dotan Cohen http://gibberish.co.il http://what-is-what.com ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
[JOB] DevoOps and software engineers, lend me your brains!
We're a young dynamic consultancy firm with strong offerings in cutting-edge devops magic, distributed systems design and automated deployment and management, IAAS wrangling, etc. We're looking for DevOps with or without cloud experience, expecially if you're the kind of guy/gal who wears a CAP and mixes SQL with NoSQL juice for breakfast. Send me your resume at ira(at)fewbytes.com. Thanks, Ira. -- I did it for my people! Ira Abramov http://ira.abramov.org/email/ ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
Re: Heterogenous, dynamic RAID for a home server
The Wikipedia article about drobo explains a bit about their disk layout. Also what about using raid 6 instead of 5? I think I read claims that it's not only more resilient nut could also be faster. On Feb 23, 2012 12:44 AM, Ira Abramov lists-linux...@ira.abramov.org wrote: Hey people, A friend of mine want to build a file server machine at home to get the benefits of a Drobo, but for 1500 NIS rather than $1500. The idea is to mix in 1T, 2T and 3T disks he already has, and on occasion grow the ׁ•torage by another disk or replace and rebalance it. He wants it to be more reliable than RAID0 but more efficient than RAID10, so RAID 5/6 are more like it. I looked around at the idea of ZFS or a distributed FS on a single node, but they are not options. I'm thinking how to do it with a block redundancy scheme, and so far I came up with this: create a RAID5 from the lower 1TB of all the disks you have, then a smaller RAID5 array from the second TB, and finally a RAID1 or RAID5 from whatever you have left. so if we say I have 2*3T and 2*2T and 1*1T, I have (5-1)*1T+(4-1)*1T+1T (the last is RAID1). now I have 8TB in three block devices, and I make a single LV (LVM2) from three extents. Is there a better solution? do note that it's not optimal for spindle activity, but this is a home machine serving two other computers and a TV streamer at the worst case. Any feedback is welcome... -- Perishable item Ira Abramov http://ira.abramov.org/email/ ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
RE: Heterogenous, dynamic RAID for a home server
I would try: 4*2T = (raid 5) net 6T 3*1T = (raid 5) net 2T Total net disk space is the same 8T but all disks are in raid Shahar -Original Message- From: linux-il-boun...@cs.huji.ac.il [mailto:linux-il-boun...@cs.huji.ac.il] On Behalf Of Ira Abramov Sent: Wednesday, February 22, 2012 3:33 PM To: IGLU Mailing list Subject: Heterogenous, dynamic RAID for a home server Hey people, A friend of mine want to build a file server machine at home to get the benefits of a Drobo, but for 1500 NIS rather than $1500. The idea is to mix in 1T, 2T and 3T disks he already has, and on occasion grow the ׁ•torage by another disk or replace and rebalance it. He wants it to be more reliable than RAID0 but more efficient than RAID10, so RAID 5/6 are more like it. I looked around at the idea of ZFS or a distributed FS on a single node, but they are not options. I'm thinking how to do it with a block redundancy scheme, and so far I came up with this: create a RAID5 from the lower 1TB of all the disks you have, then a smaller RAID5 array from the second TB, and finally a RAID1 or RAID5 from whatever you have left. so if we say I have 2*3T and 2*2T and 1*1T, I have (5-1)*1T+(4-1)*1T+1T (the last is RAID1). now I have 8TB in three block devices, and I make a single LV (LVM2) from three extents. Is there a better solution? do note that it's not optimal for spindle activity, but this is a home machine serving two other computers and a TV streamer at the worst case. Any feedback is welcome... -- Perishable item Ira Abramov http://ira.abramov.org/email/ ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il ___ Linux-il mailing list Linux-il@cs.huji.ac.il http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il