0

I've got 4 1TB Hard-Drives. I want to have ONE mounted directory in which there are 2 directories. So the space for both is dynamic. One needs a parity, the other one not. Of course I don't want to store the parity on the same disk as the original data.

I would preferably use Ubuntu-Server as operating system. Does anyone know a software or OS the realize this?

Ma_124
  • 3
  • 2
  • One way to get copies of your data is to use ZFS and make a ZFS filesystem which has `copies` attribute set to two on your filesystem that contains the more important data. However, this does not guarantee that the data will be stored on both devices. – Tero Kilkanen Jun 07 '17 at 21:55

1 Answers1

0

You have much more to learn on this subject than finding a software or operating system. But as everyone has to start from somewhere, here is a short briefing with links for more information.

I want to have ONE mounted directory in which there are 2 directories.

In Unix/Linux you have a single namespace /*, a big file hierarchy tree where files can be spread across several devices: not only the hard drives but also the memory ( /proc for process information pseudo-file system) and devices (/dev, location of special or device files). In Linux Filesystem Hierarchy everything is a file. In this manner you can mount different drives or partitions as subfolders in a single folder, e.g. /path/to/folder/partition1 and /path/to/folder/partition1. Possible.

So the space for both is dynamic.

No. Partitions aren't dynamic. If you use separated partitions, you must decide fixed size for them.

One needs a parity, the other one not. Of course I don't want to store the parity on the same disk as the original data.

There's a huge terminological difference between parity and mirroring. If you need parity data, you need at least three drives: two of them has the actual data and one counts them together for parity data (RAID 3, RAID 4). If the parity data drive fails, new parity data must be generated: if any of the data drives fails, it can be regenerated using data from the other drives along with the parity data.

Mirroring (RAID 1) creates an exact copy of the drive. It needs two drives, has good performance and excellent redundancy. Striping (RAID 0) spreads data across drives without any redundancy or fault tolerance, but is good if the good performance is the only goal.

With all RAID configurations you use the entire drive and either do mirroring/striping/parity or not. With your four identical drives you have many options. You could create RAID 4 with three drives and use the fourth drive without redundancy, which would be as close as it gets to your original specification. However, you won't gain more space and certainly not more redundancy and performance compared to the way superior choice: build nested RAID 10 using all the 4 drives; 2 TB of fault tolerant storage.

RAID 10 is the combination of RAID 0 and RAID 1. It applies parity check to realize striping set of mirror. So it inherits speediness of RAID 0 and security of RAID 1.

Esa Jokinen
  • 46,944
  • 3
  • 83
  • 129