RAID is a great system for increasing speed and availability of data as it offers considerably more data protection than non-RAID disk systems. However, its management of the disks and the data distribution across them can be extremely complex.
Vogon has extensive experience of recovery from RAID, spanned and striped systems. We only normally require just the storage devices in order to recover lost data. If you are asked for the original RAID or array system by a Data Recovery company, you must ask yourself whether that company really understands how the system works.
Different RAID levels are utilised for different applications, depending upon the fault tolerance, speed of access required or the average size of the files being stored.
RAID systems do fail!
Not something that a RAID vendor would want to admit to, but complex redundant systems can suffer from failure. Often not a fault of the technology used or the design of the array, but typically failure to correctly implement these systems leads to a single point of failure that can cause catastrophic data loss.
No matter how well designed or implemented the system is, there is still one very complex factor that causes most RAID array problems we see – the human being. Mistakes are easy to make and the more complex the systems become, the more potential there is for mistakes to occur.
Vogon has an excellent working relationship with all of the major peripheral, software and media manufacturers and has developed custom software and firmware for major American and Japanese peripheral manufacturers and software companies.
- Multiple drives can fail in an array – it's improbable, but bad luck does strike
- Arrays are normally boxed in a single case, so physical damage can affect multiple drives and the control electronics
- Many people don't back up RAID systems because they're 'fault tolerant' – however they're not 'fault proof'
Vogon RAID and array recovery processes
Using our disk recovery processes, coupled with our ability to produce a safe 'copy' of the complete volume, allows us to process an array as a collection of image files. Our Data Recovery systems have the capacity to absorb most server array volumes.
A project team of Data Recovery Hardware Specialists and Data Recovery Software Specialists is formed to deal with each recovery. While the failed drives are recovered onto the Data Recovery servers by the hardware team, the software team evaluate the RAID or array images, work out the configuration and determine the extent of the corruption that typically occurs when these systems fail in operation.
The advanced software tools we have written will extract the data from the images. When a drive image is not available, the tools can reconstruct the data 'on-the-fly' in the same way that the RAID rebuild process would have done on the original system.
At all stages, controlling the tools is an experienced software expert who closely understands the software tools, the array configuration, the file system on the array, what the problem is and why the array failed in the first place.
The data will be returned using the customer's chosen media.
What are the RAID levels?
RAID 0 – Striped drives without parity, a non-redundant group
This is the fastest and most storage efficient configuration of a group of disks, but it has no fault tolerance; if one drive fails the whole array fails. As there is no parity, it offers the best storage efficiency. The striping allows read/write operations to occur simultaneously on each disk for speed.
RAID 1 – Mirroring drives
The fastest fault tolerant array configuration but the least storage efficient array, it's probably the most commonly used system today. As other array configurations require three or more drives, it is the only choice in a two-drive system. The two drives in the 'mirrored pair' appear to the operating system as one; they always mirror each other. Should a drive fail the data is available from the other. As the data is always the same on the two drives, read performance is almost doubled, as consecutive reads can occur on each drive at the same time. Write performance is not improved.
RAID 2 – Sector striping with a disk holding Error Correction Code (ECC) information
As all hard disks have ECC data built into every sector, this configuration has no advantages over RAID 3 and is not used.
RAID 3 – Sector striping with a disk holding parity information
A single disk holds the parity information and the data integrity checking relies on ECC on the disk drives themselves, as with RAID 2. Because parity information is held on a dedicated drive, any single drive in the array can fail without data being lost. Data can be rebuilt onto a replacement drive using the remaining information and the parity information. To maintain performance when transferring small files, synchronised motors may be required on each disk. Read and write operations cannot be overlapped, but as consecutive reads can occur on each drive at the same time, read performance is improved.
RAID 4 – Very large sector striping with a disk holding parity information
This is a faster version of RAID 3, in that the complete read of a file can occur on a single disk. Write performance is not improved as the parity drive needs updating with every write operation. As there are no real benefits over RAID 3 and write performance is worse than RAID 5, this configuration is not used.
RAID 5 – Large sector striping data with a rotating parity stripe
With no dedicated parity drive the write performance is better than RAID 3, with overlapped data and parity update writes. Read operations can be interleaved as all drives contain data stripes. RAID 1 performs faster but RAID 5 provides better storage efficiency. Parity update can be more efficiently handled by RAID 5, by checking for data bit changes and only changing the corresponding parity bits. For small data writes improvements here are lost as most disk drives update sectors entirely for any write operation. For larger writes only the sectors where bit changes need to be made are rewritten and improvements made. Maintaining parity information reduces write performance – in some cases as low as one third the speed of RAID 1. For this reason RAID 5 is not normally used in performance critical processes.
RAID accomplishes the formation of a single logical drive from multiple units through a process called striping. Striping involves logically arranging information so that individual files are spread among a group of drives. A stripe segment can be as small as a single byte, or as large as multiple sectors. Striping has the advantage of faster data access because individual drives can be accessed in parallel. The disadvantage is that the array capacity, once formatted, is fixed, meaning expansion is not as simple as adding drive modules.