[prev in list] [next in list] [prev in thread] [next in thread] 

List:       zfs-discuss
Subject:    Re: [zfs-discuss] ZFS + OpenSolaris for home NAS?
From:       Peter Korn <peter.korn () sun ! com>
Date:       2009-01-07 2:29:19
Message-ID: 640612041.271231324189766.JavaMail.Twebapp () sf-app1
[Download RAW message or body]

I'm looking to do the same thing - home NAS with ZFS.

I'm debating several routes/options, and I'd appreciate opinions from folks here.

My system will primarily be a file & music server, serving CIFS and some NFS as well \
as driving multiple concurrent audio streams via SqueezeCenter, and perhaps \
eventually videos via something like MythTV.  I'm also contemplating making it a Sun \
Ray server.  I don't want to run out of disk space for a long time - I'd like >= 3TB \
to start.

Decision #1: AMD vs. Intel
I'm drawn to quad core, perhaps for silly reasons.  I'm also drawn to AMD (for price \
and RAM performance).  And finally I'm drawn to ASUS motherboards because there are \
                so many options in the HCL.  I want lots of SATA ports and support \
                for 8GB RAM
 - option a: ASUS M3A-H/HDMI ($56@Newegg) with AMD Phenom 9600 ($119@Newegg)
 - option b: ASUS P5K Deluxe ($169@Newegg) with Intel Quad Q6600 ($189@Newegg)
Any reason to go with Intel?  Any better option going the AMD route (perhaps the Quad \
core chip offers no value?) 

Decision #2: 1.5TB Seagate vs. 1TB WD (or someone else)
I've heard lots of stories about Seagate reliability being poor, especially in RAID \
configurations.  WD has a good reputation, especially with their more expensive E3 \
drives.  Reliability is a concern both from a time point of view (spending time \
getting/installing replacements and downtime until it is in), as well as potential \
                data reliability if a multiple-failure occurs.
 - option a: 3 1.5TB Seagate drives (in Raid-Z) or 4 1.5TB in Raid-Z2 ($129@Newegg)
 - option b: 4 1TB WD drives (in Raid-Z) or 5 in Raid-Z2
  - option b(1): WD RE3 drives ($150 via eBay seller)
  - option b(2): WD Caviar Black ($119@Newegg)
  - option b(3): WD Caviar Green ($119@Newegg)
Do I loose much in performance going Green vs. Black?  Do I gain anything in \
reliability?  Anything measurable (given other system components) in power savings?

Decision #3: Raid-Z vs. Raid-Z2
I understand that with very large drives, the chance of a momentary block read \
failure is near (or in excess) of the 1/# blocks on disk (so you would likely get one \
of these if reading/copying the entire 1-1.5TB), and while subsequent re-tries would \
fix this it would be seen as a RAID failure in the moment.  But the more I read up on \
ZFS, the less likely a double-fault situation appears that it'd occur, given the \
                fix-on-error-detection and continuous cleaning of the disk.
 - option a: Raid-Z
 - option b: Raid-Z2

Decision #4: file system layout
I'd like to have ZFS root mirrored.  Do we simply use a portion of the existing disks \
for this, or add two disks just for root?  Use USB-2 flash as those 2 disks?  And \
                where does swap go?  
 - option a: [3 disks]
  - option a(1): reserve ~50GB from all three, ZFS mirror across all three, and put \
                swap into that 50GB
  - option a(2): reserve ~50GB from all three, ZFS mirror across first two, and put \
                swap into the third 50GB (rather large that...)
  - option a(3): reserve ~50GB from all three, ZFS mirror across first two, and put \
swap into that 50GB; ignore the 50GB in the third

 - option b: [4 disks]
  - option b(1): reserve ~25GB from all four, ZFS pool/mirror across all four, and \
                put swap into that pooled 50GB
  - option b(2): reserve ~50GB from all four, ZFS mirror across first two, and put \
                swap into the third 50GB (rather large that...) and ignore the fourth \
                50GB
  - option b(3): reserve ~50GB from all four, ZFS mirror across first two, and put \
                swap into the third 50GB (rather large that...) and ignore the fourth \
                50GB
  - option b(4): reserve ~50GB from all four, ZFS mirror across first two, and put \
swap into that 50GB (rather large that...); ignore the third and fourth 50GB

 - option c: [5 disks]
  - option c(n): you get the idea...

 - option d: [n disks + 2 "root" disks]
  - option d(1) use some small cheap pair of SATA (or even IDE) disks for mirrored \
                root/swap
  - option d(2) use ~8GB USB2 flash drives for mirrored root/swap (and place things \
like /opt into the larger "data" array)


Many thanks for your feedback!

Peter
-- 
This message posted from opensolaris.org
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss


[prev in list] [next in list] [prev in thread] [next in thread]