A-Z of NAS and RAID Data Recovery

BASIC

A proxy term used by NAS manufacturer Synology to denote JBOD.

BeyondRAID

This is a proprietary type of RAID developed by Data Robotics (now Drobo) and used on Drobo NAS devices. A hybrid of RAID 1 and RAID 5, it allows users to mix-and-match drives of different sizes. This file system cannot be mounted or read by other file systems.

Btrfs

A next-generation file system, whose many features include self-healing, check-summing of metadata, automatic compression and fast snapshots. Btrfs can recover from broken B-tree roots when the volume is remounted. The Btrfs-restore command can be used for data recovery from un-mountable Btrfs volumes. As a file system, Btrfs sounds too good to be true. However, this is still in the experimental stages of development and it will probably take another 5 years or so until it's kernel has matured to a stable format.

CCTL

(Command Completion Time Limit) - The name used by Samsung and Hitachi for their version of Error Recovery Control.

DAS

(Direct Attached Storage) - Usually a RAID 0 or RAID 1 device with no redundancy. It is attached directly to the host computer system, via USB or a Thunderbolt connection. It is the simplest way to get speed and volume.

Discrete RAID Controller Card

A plug-in expansion card which has a built-in RAID processor, and usually plugs into one of the PCI slots on the host system's motherboard. In this type of set-up, the RAID is independent of the host system's processor. As a result, this type of controller allows for high performing I/O cycles and better memory performance.

EXT2

A file system developed by Remy Card in 1993. When first introduced, EXT2 had a number of innovative features. For example, the user was allowed to choose the optimal block size (from 1024 to 4096 bytes). It was, at the time, one of the few file systems that allowed the user to choose how many inodes were allowed for a partition of a given size. It also introduced file system checks using e2fsck. This function could be executed after a system crash, as well as after a set number of file system mountings.

EXT3

Developed by Stephen Tweedie in 2001, EXT3 is an improved version of EXT2 that supports journaling. It has a maximum block size of 4096 bytes and a special zone on the file system that is reserved for journaling. By tracking every change in the file system and the individual files, it improves the probability of a successful data recovery from EXT3, should the volume become corrupted.

EXT4

Introduced in 2008, it can handle individual file sizes of up to 16TB. The file system itself can support disks up to 1 exabyte. As extents are used instead of inodes, similar file types are stored contiguously. One extent can map up to 12MB of space if a 4KB block size is used. Users can mount an existing EXT2 or EXT3 file system to EXT4 without having to upgrade it. (This upgrade-ability of EXT file systems was a key reason why the search engine giant Google chose EXT4 for it's storage infrastructure). The file system is currently the default file system used by Synology NAS devices.

Extent

this is a single descriptor that represents a range of contiguous physical blocks. Extents are not very effective for representing sparse or non-contiguous files, but they do make the storage of contiguous files very efficient.

Error Recovery Control

(ECR) - A feature of modern hard disks which allows the user to specify the amount of time which a drive's firmware is allowed to either recover or correct a read / write error. On enterprise-class disks, this feature is often enabled by default. ERC will usually allow a drive's firmware a set period of time (e.g. 7 or 8 seconds) to process a read or write error, which prevents the disk from being marked as "failed" by the RAID controller. See also TLER

Firmware RAID

(also referred to as BIOS RAID or motherboard RAID) - A type of RAID where the RAID configuration is stored on a non-volatile memory chip on a printed circuit board. In the case of desktop / server computer environments, this is usually an NVRAM chip on the motherboard. The problem with this type of RAID is that any motherboard failure (apart from RAID1) will also result in a failed array.

FreeNAS

Open source software RAID that is based on FreeBSD and the ZFS file system. FreeNAS is commonly used on old desktop PCs that have been re-purposed as NAS servers.

Hardware RAID

A hardware RAID uses its own processor and memory to run the RAID application. It can be either integrated into the motherboard of the host system or it takes the form of an add-in card. Typical add-in RAID cards include those from LSI and Adaptec.

Inode

This stores basic metadata information about a file, such as user access control, timestamps and pointers to data blocks. See also Extents

JBOD

The power of JBOD lies in its ability to take physical drives and combine them into one large logical drive that is presented to the host operating system.

Journaling

A process where file system changes are tracked and recorded in a journal log. Journaling can be a useful function to reduce the impact of system crashes and power failures, as it prevents the file system from becoming inconsistent. For modern file systems, like EXT4 and modern NTFS, sophisticated journaling has become their killer feature.

Latent Error

A latent error occurs when a damaged sector on a disk gets written to, and it is not detected by, the file system.

Online Capacity Expansion (OCE)

This offers users the ability to increase the storage capacity of an existing RAID array, usually by adding one or more disks. This can be a very useful feature on RAID arrays, but does not come without risks. If corruption occurs during the expansion process, for whatever reason, you risk your RAID volume becoming inaccessible. It is therefore strongly recommended that your NAS or server is fully backed-up before you execute online capacity expansion.

Parity

A data error checking method that uses an extra bit in which the total number of binary 1's (or 0's) in a byte is always odd or always even. Thus, in an odd parity scheme, every byte has eight bits of data and one parity bit. If you are using odd parity and the number of 1 bits that comprise the byte of data is not odd, the 8th or parity bit is set to 1 to create odd parity. Using this technique, a byte of data can be checked for accurate transmission by simply counting the bits for an odd parity indication. If the count is ever even, it is usually indicative of an error in the array.

Parity Data

A block of information created from several blocks of user data to allow for data recovery from an array which has failed.

Mdadm

A useful Linux utility to manage devices on a Linux software RAID. It allows the user to create, manage and monitor software RAID arrays.

ReiserFS

File system developed by Namesys that is no longer commonly used by any major distribution. It is, however, the default for a small number of distributions, including Xandros and Linspire.

ROC

(Raid on Chip) - A type of hardware RAID where the RAID controller is stored on a separate processor chip. If there is no processor, the RAID is usually referred to as a software RAID.

Scrubbing

Sometimes sectors on a disk will be damaged, but still get written to. This is known as a latent error. In RAID arrays, latent errors can lead to disk failure. Disk scrubbing is a background process that checks sectors during idle periods. If a defective sector is found, it can be rewritten or remapped to a used part of the disk(s). In the context of RAID data recovery, sector remapping is not always desirable as remapped data is often not contiguous and can result in corrupted data.

SHR

Synology Hybrid RAID is the proprietary RAID used in some Synology NAS devices. Like other proprietary RAID systems, it offers a greater level of flexibility than conventional RAID. Like BeyondRAID, it allows users to mix and match drives of different sizes without leaving any wasted space.

Software RAID

Software RAID uses the host system's CPU for processing power. Like all software, the software used to manage the array can go corrupt, leading to an inoperable storage device and inaccessible data.

SMB (Server Message Block)

File writing protocol developed by IBM, which facilitates data transfer between client applications and a host server.

SMB2

This is an improved version of the SMB file sharing protocol. Window scaling and WAN acceleration have been introduced. Macs running OS X 10.9 (and above) will use SMB2 as their default file sharing protocol when connecting to other Macs, Windows PCs / servers and NAS devices.

SMB3

With the prevalence of virtual data servers, Microsoft introduced a number of performance enhancements for SMB3, such as multichannel and transparent failover.

Striping

The process of breaking down a file into smaller pieces. When data is passed through a RAID controller, it is divided by the pre-determined stripe size to create 1 or more blocks. These blocks are then spread across the drives in the array, which makes reading and writing to a volume faster.

Stripe Size

This is the size of the entire stripe spread across all the disks in a RAID disk group. Stripe size can be calculated as follows: Stripe Unit Size X Numbers of Disks in RAID groups. In the case of a failed RAID array, the calculation of the exact stripe size used is critical to the success of the data recovery process.

Stripe Unit Size

This is the size of each element of the stripe, as stored on each member of the disk group.

TLER (Time-Limited Error Recovery)

This is the Error-Recovery Control feature used by Western Digital drives. Most modern hard disks will try to correct the algorithmically small read or write errors that are processed by the drive's MCU. This can take several seconds to complete, during which time the drive might seem to be unresponsive. If the drive is being used in a RAID setup, the RAID controller might see the unresponsive drive as "faulty" and drop it out of the array. To counteract this, all Western Digital enterprise-class and RAID drives have TLER enabled by default. TLER will interrupt error correction, in case RAID perceives the disk as faulty. This is a very useful feature, especially if your RAID controller card is sensitive to hard drive read / write delays.

Windows RAID

Microsoft's operating systems (e.g. Windows 7 and 8) allow users to create a software RAID. Microsoft offers three different types of RAID variations, including spanned, striped and mirrored.

Write Ahead Logging

see Journaling

XFS

File system developed by Silicon Graphics, Inc. to support large files, e.g. video files on larger partitions. XFS is scalable and fast. However, when used to store lots of small files, e.g. document files, it can be painstakingly slow. When using XFS, it is strongly recommended that your server or NAS uses ECC memory and the host system uses a UPS device. (XFS does not handle sudden power loss very well).

XOR

The function used to calculate data bit relationships in a RAID array. Data recovery from RAID relies on the Boolean logic of XOR to reconstruct a corrupted array. XOR analyses a series of 0's and 1's resulting either "true" or "false". Even numbers are designated "true", while odd numbers are "false". By using this logic, the controller can fill in the blanks - just like in algebra. We know that 2+3=5. If we see an equation like 2+x=5, we know that ‘x' is meant to be a 3. XOR uses the same principle for maintaining RAID integrity.

X-RAID

A proprietary type of RAID used by Netgear ReadyNAS devices. It is a variant of the standard RAID 5, which is based on the EXT2 / EXT3 file system. Like many other rival proprietary RAID systems, it is auto-expandable. In other words, you do not have to rebuild the array to add more capacity.

ZFS (Zettabyte File System)

This is a 128-bit file system developed by Sun Microsystems in 2005 that supports compression (LZJB and gzip) and snapshots. The real killer feature of ZFS is its superb journaling feature that helps to ensure the file system remains consistent in the event of bit errors or power loss. However, ZFS is not commonly found on NAS devices, due to it's high CPU overhead.