[ale] ZFS on Linux

Derek Atkins warlord at MIT.EDU
Mon Apr 22 13:24:44 EDT 2013


Jim Kinney <jim.kinney at gmail.com> writes:

> given what the drive arrays are used for, dedup is a big plus (people make
> copies and copies of research files like drive space is $0!). And, yeah, even
> with no dedup, 32G is not going to cut it anyway.

You can get mobos that support 256MB for not too much money.

For my case I cannot imagine that dedup would significantly help me; my
plan is to store audio and video, and perhaps *some* VM data, but most
likely I don't imagine that dedup would do *me* any signigicant
advantage.

> The checksums for RAID6 are the same as in raidz2.

Are they?  I though ZFS/RAID-Z(2) had additional checksums above and
beyond what RAID5(6) provided, such that it could detect and correct a
write-hole (or other) disk/storage failure.

> I found a 250G SSD but can't use it. I have one other system I can put it on
> but it has not enough PCIe slots for the need.
>
> the solution to all of this is $$$$$ for new system :-(

Or just replace the mobo ;)

-derek

> On Fri, Apr 19, 2013 at 11:58 AM, Derek Atkins <warlord at mit.edu> wrote:
>
>     Jim,
>    
>     Jim Kinney <jim.kinney at gmail.com> writes:
>    
>     > After much research, I have decided to NOT implement ZFS. The main
>     factors of
>     > interest were the raidz2 and data deduplication. In order to support the
>     > deduplication,the server needs 5G RAM per 1TB storage. That's a deal
>     breaker
>     > right there as the system to be used will only support 32G and has 50TB.
>     > Raidz2 is basically software RAID6 which I can already do.
>     >
>     > Looked at getting an SSD for the deduplication data and using l2arc but
>     I have
>     > no space to install it on that server. Grr.
>    
>     Do you really feel you need ZFS deduplication?  According to
>     http://doc.freenas.org/index.php/Hardware_Recommendations you only need
>     5GB/TB if you're using dedup.  If you're just using ZFS then they
>     recommend 1GB/TB.  Granted, if you've got 50TB of data then a 32GB
>     memory limit wont be useful to you.
>    
>     However, thank you for making me look at that; I was considering a
>     32GB-limit Mobo for my system, which I was expecting to start at 12TB
>     usable but potentially expanding to 96TB.  I don't think a 96TB ZFS
>     would work well with 32GB memory.
>    
>     OTOH, is ZFS really only just "RAID6"?  I thought that the ZFS checksums
>     would provide improved data integrity (scrubbing) over pure RAID6?
>    
>     -derek
>    
>     > On Fri, Mar 29, 2013 at 3:10 PM, Ed Cashin <ecashin at noserose.net> wrote:
>     >
>     >     Thanks in advance for the notes you take.  I'm interested to hear
>     what you
>     >     find.
>     >
>     >     By the way, work sent me to a Nexenta course in 2010 where I learned
>     ZFS
>     >     stuff from Richard Elling:
>     >
>     >       http://www.richardelling.com/
>     >
>     >     ... and one thing that kept coming up is how users usually reach for
>     the
>     >     RAID-Z options when stripes of mirrors would perform better and
>     offer more
>     >     flexibility.  I thought that was notable.
>     >
>     >     On Fri, Mar 29, 2013 at 2:44 PM, Jim Kinney <jim.kinney at gmail.com>
>     wrote:
>     >
>     >         There was an update to this that just hit /.
>     >
>     >         WooHoo!
>     >
>     >         I have not used ZFS (minimal solaris experience) before but I
>     have a
>     >         situation where raidz looks like a very good choice. An external
>     16
>     >         bay JBOD box hanging off a SAS HBA looks like a prime candidate.
>     The
>     >         big issue for me is the support for native NFSv4 ACLs which
>     solves
>     >         some issues.
>     >
>     >         As I go through this exercise, I'll keep some notes and maybe
>     look at
>     >         a presentation on doing this.
>     >
>     >         hints, gotcha's, are very welcome.
>     >
>     >         I did find a pdf of the zfsadmin.pdf  http://pdfhome.org/
>     zfsadmin.pdf
>     >
>     >         --
>     >         --
>     >         James P. Kinney III
>     >
>     >         Every time you stop a school, you will have to build a jail.
>     What you
>     >         gain at one end you lose at the other. It's like feeding a dog
>     on his
>     >         own tail. It won't fatten the dog.
>     >         - Speech 11/23/1900 Mark Twain
>     >
>     >         http://electjimkinney.org
>     >         http://heretothereideas.blogspot.com/
>     >
>     >         _______________________________________________
>     >         Ale mailing list
>     >         Ale at ale.org
>     >         http://mail.ale.org/mailman/listinfo/ale
>     >         See JOBS, ANNOUNCE and SCHOOLS lists at
>     >         http://mail.ale.org/mailman/listinfo
>     >
>     >     --
>     >       Ed Cashin <ecashin at noserose.net>
>     >       http://noserose.net/e/
>     >       http://www.coraid.com/
>     >
>     >     _______________________________________________
>     >     Ale mailing list
>     >     Ale at ale.org
>     >     http://mail.ale.org/mailman/listinfo/ale
>     >     See JOBS, ANNOUNCE and SCHOOLS lists at
>     >     http://mail.ale.org/mailman/listinfo
>     >
>     > --
>     > --
>     > James P. Kinney III
>     >
>     > Every time you stop a school, you will have to build a jail. What you
>     gain at
>     > one end you lose at the other. It's like feeding a dog on his own tail.
>     It
>     > won't fatten the dog.
>     > - Speech 11/23/1900 Mark Twain
>     >
>     > http://electjimkinney.org
>     > http://heretothereideas.blogspot.com/
>     >
>     > _______________________________________________
>     > Ale mailing list
>     > Ale at ale.org
>     > http://mail.ale.org/mailman/listinfo/ale
>     > See JOBS, ANNOUNCE and SCHOOLS lists at
>     > http://mail.ale.org/mailman/listinfo
>    
>     --
>            Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
>            Member, MIT Student Information Processing Board  (SIPB)
>            URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
>            warlord at MIT.EDU                        PGP key available
>    
>     _______________________________________________
>     Ale mailing list
>     Ale at ale.org
>     http://mail.ale.org/mailman/listinfo/ale
>     See JOBS, ANNOUNCE and SCHOOLS lists at
>     http://mail.ale.org/mailman/listinfo
>
> --
> --
> James P. Kinney III
>
> Every time you stop a school, you will have to build a jail. What you gain at
> one end you lose at the other. It's like feeding a dog on his own tail. It
> won't fatten the dog.
> - Speech 11/23/1900 Mark Twain
>
> http://electjimkinney.org
> http://heretothereideas.blogspot.com/
>
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://mail.ale.org/mailman/listinfo/ale
> See JOBS, ANNOUNCE and SCHOOLS lists at
> http://mail.ale.org/mailman/listinfo

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord at MIT.EDU                        PGP key available



More information about the Ale mailing list