This tool calculates the total available storage and reliability information for different RAID and disk configurations.
Here are some typical settings for an analysis of a JBOD with 45 bays where the RAID configuration is Z2, each RAID array consists of 6 disks and there are 3 left over hot spares (45 - 7*6 = 3).
RAID Option | RAID-6/Z2 |
MTBF | 750,000 |
MTTR | 24 |
Capacity | 2 |
Disks per Array | 6 |
Arrays per JBOD | 7 |
Spares | 3 |
ZFS Compression Factor | 1.3 |
Definitions
Array | A collection of disks in a RAID configuration. |
Capacity | The disk capacity in TBs. A 500GB would be specifed as 0.5. |
Disks per Array | The number of disks in each RAID array or virtual device (vdev). Normally the disks are organized as a collection of vdevs where each vdev has 2 or more disks. Some RAID configurations require a minimum number of disks in each vdev such as RAID-10 which requires a minimum of 2. The size of each vdev in a JBOD must be the same. All RAID configurations are limited by the size of the smallest disk (least capacity). |
JBOD | Just a bunch of disks. It is normally a chassis that can hold a fairly large number of disks that are connects via SATA or SAS to a controller. Typical sizes are 8, 12, 24, 36 and 45. A JBOD can be thought of as a container for vdevs and hot spares. |
MTBF | Mean time between failures for a single disk in hours. This is typically 750,000 hours or a little more than 85 years. |
MTTDL | Mean time to data loss as described in this paper. |
MTTR | Mean time to repair a single disk in hours. This is how long it takes to physically replace a bad disk. It is typically about 24 hours and is important because other disks can fail while waiting for the repair to take place. |
RAID | See wikipedia for an in-depth explanation of the different RAID configurations. |
Spares | The number of disks that can automatically be used as hot spares. When a disk fails in a vdev, ZFS has the ability to automatically replace it with a hot spare until it can be replaced. A hot spare is a spare that can be replaced while the system is running. This is required for systems where you cannot afford to shutdown the storage to replace a single disk. |
Spare Tolerance Factor | In ZFS, hot spares will automatically take over for any failing disk which means that they can act like an additional parity stripe to improve performance. When this factor is greater than 1, it is treated like an additional disk in each array. When it is less than 1 but greater than 0 you will see a > sign to indicate that the reliability is higher than if it wasn't there but without providing a magnitude. The reason for that is that I do not know how to calculate the increased reliability because there is no notion of partial parity stripes in the literature. |
Vdevs | A virtual device is a collection of 2 or more physical disks that are treated as a single device in the file system. They are not treated separately buy the controller. |
ZFS Compression Factor | The additional data capacity provided by ZFS compression. No compression is specified as 1. A typical value is 1.3 which means that 100TB of raw data capacity would yield 130TB of data when compression is enabled. This is such a common ZFS configuration that it really needed to be here. Dedup is not quite as common so I didn't include it but it can also really improve capacity for the right datasets. |