Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


software raid
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

software raid

camargcamarg Member
edited December 2011 in General

Just got a dedi with 7 x 3TB hdd to use for backups.
I'm between raid5 with 1 spare & raid6.
Which one would be better? I'm more concerned about data integrity.

Comments

  • KuJoeKuJoe Member, Host Rep

    RAID6, you can lose 2 drives without data loss thus more redundancy so if you lose 1 drive, you can throw in a replacement and while it's rebuilding you'll still have another parity drive.

    RAID5 with 1 spare allows you to lose 1 drive without data loss but you're without any protection while your spare drive rebuilds (very slowly with software RAID).

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    The problem i've seen with software raid6 is that the write performance is just over a single drive. While this is supposed to be how 'real raid 6 is', most RAID6 setups have a large cache to help it along :)

    You can tune mdadm to give you a ~128MB cache on linux, but it's only so good.

    @mitgib has a decent sized software raid6 so he can chime in if he isn't too busy trollin'

    Francisco

  • In a software raid, the need for 14 parity stripes over 7 drives instead of 7 parity stripes over 7 drives may make throughput low and/or machine load high. However, this is a backup box, so throughput is not that important.

    I'd also choose RAID 6, especially since this is a backup box.

  • jhjh Member

    With 7 big drives I would recommend a raid card. Is this not an option?

  • @jtodd it's an xs13 from hetzner, they don't provide raid controller for this setup.
    I'll probably go with a raid6 setup.

    anybody knows a gimmick to make partitions larger than 2TB on centos installation?

  • DamianDamian Member
    edited December 2011

    @camarg said: anybody knows a gimmick to make partitions larger than 2TB on centos installation?

    When we ran into this on Centos 5, we ended up making a 10gb / partition, then when the system was installed, we made the rest of the RAID space into the /vz partition. In this case, you could make a /backup partition instead, unless your backup scheme warrants your directory tree to be part of the / file system.

    I thought Centos 6 was able to install to whatever the maximum partition size is for your selected file system?

  • jhjh Member

    I was under the impression only the boot partition had to be <2TB

  • @camarg said: anybody knows a gimmick to make partitions larger than 2TB on centos installation?

    Use GPT, except for the boot drive. And if all you have is 7 3tb drives, you are fscked cuz the boot drive needs to be msdos, you can't boot from GPT except EFI and I'm sure you have a bios. Just a USB thumb drive would be enough so you can install the boot blocks on it.

  • What is software RAID-6 performance like? I would assume it's a bit worse than RAID-5 due to extra parity. Personally I wouldn't do software RAID-5/6 at all, but curious about the performance hit. Anyone using it?

  • KuJoeKuJoe Member, Host Rep

    @qenox said: What is software RAID-6 performance like?

    For backups I don't think it matters. Unless you have deadline, I don't think it would make a difference if it took 17 hours or 1 hour to backup your data. Even if it only wrote at 15MB/s, that's still faster than a 100Mbps port can download so in most cases you're ok. If you're using rsync then you'll only need to backup all of the data once and the after that it doesn't matter how slow the server is.

    Thanked by 1japon
  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @qenox - for writes you'll get dd speeds equal to that of 1 disk. For reads it's a lot faster.

    Francisco

  • Thanks KuJoe @ Francisco.

    @Francisco, I'm surprised, you would expect the parity calculation to have a noticable effect on the CPU, but I guess CPUs nowadays are quite powerful so the hit isn't as much.

    I guess I'll still stick with HW raid though for RAID-5/6. @KuJoe, yeah, we do have a dead line for our backups, so HW raid it is.

  • KuJoeKuJoe Member, Host Rep
    edited December 2011

    We have a HW RAID5 setup for our Backup VPS clients and we were showing 250MB/s write speeds before switching to the OpenVZ kernel (dropped to 200MB/s).

Sign In or Register to comment.