Understanding RAID Systems for Surveillance Storage

A tutorial on RAID data storage architectures, and how they apply to video surveillance

So is RAID 1 the answer for physical security systems? In many cases, the answer is no. That's because by doubling the disks used to record a mirror, you effectively also double the cost per megabyte of storage. And that can add up very quickly with video workloads requiring longer retention and higher frame rates and better resolution, to the point that DVRs quickly run out of recording capacity.

While RAID 1 is undoubtedly a useful technology for some applications, it is clearly even better financially for those selling disk drives, who will sell double the capacity for every storage application using it compared with those that don't. That is why we would say that RAID 1 should not typically be your first choice.

We don't have to spend a lot of time on RAID 2, as it is not used in any commercial storage implementation today and you can skip ahead. But if you just have to know why, read on.

When the RAID definitions were originally laid out, RAID 2 made sense because it striped data to a bunch of drives, with an ECC (Error Checking and Correction) code symbol stored on a specific drive for the entire array for data recovery purposes. As ECC codes are now embedded in disk drives, however, this is no longer required and RAID 2 is no longer commercially used.

RAID 3 copies or 'stripes' data across three or more drives at the byte level. One of the drives becomes the parity drive and performs this task. The parity drive is not used to store video or other digital data. By using the information on the parity drive, it is possible to rebuild and recover all of the information on the remaining drives in the event of a disk failure.

Only in the extreme case of two drives failing at once will RAID 3 falter and stored data become lost. So while RAID 3 has some risk (if a second drive were to fail before the first drive was rebuilt), it greatly reduces the cost in comparison to RAID 1 and improves the availability of the data over having no RAID or only performance-oriented RAID 0 in use.

RAID 3 has been used in some IT systems, but due to the inherent complexity of the storage controller design needed and some configuration challenges when rebuilding lost data, it has been largely superseded by RAID 5. It is not commonly used today.

RAID 4 functions like RAID 3, but uses larger stripes so as to allow read speed to be increased. While an improvement in some environments over RAID 3, it offers no real advantage over RAID 5.

Thus like RAID 3, RAID 4 has been largely replaced by RAID 5 and is not found in most storage systems today.

We've already talked about several RAID levels giving way to RAID 5, so briefly here's what is and why it is so popular.

Simply put, RAID 5 delivers the best overall balance of data protection and performance and makes the most efficient use of drive capacity of all RAID techniques. As a result, RAID 5 is the most widely used technique in IT and is also found in the best DVR and IP storage systems.

RAID 5 requires three or more identical drives to be in a RAID group. Data is striped across drives, and parity is distributed amongst them all.

While RAID 5 is not as fast as RAID 1 since its parity data is spread across multiple drives, it is clearly the best overall choice for many physical security and IT applications.

Data loss risks are greatly lessened with RAID 5. Only in the unlikely event of a second drive failure, just as with the other RAID levels we have reviewed, will data be at risk.

RAID 6 is an extended version of RAID 5 that it is able to protect against not just one, but two simultaneous disk drive failures without loss of recorded data.

RAID 6 stripes data on a block level, just as RAID 5 does. But it also adds a second set of parity that is calculated and then copied to all of the drives in the array. This extra step is what allows RAID 6 to survive a double disk drive failure, but it also requires more disk space and thus is less efficient than RAID 5.

In particular, RAID 6 is considered to be unacceptably inefficient when a small number of low capacity disk drives are deployed, such as in a typical DVR. However, the ability of RAID 6 to support two simultaneous failures is growing in importance as the number of hard drives in individual IT storage systems dramatically increases to support larger numbers of megapixel digital cameras at higher resolution and frame rates.