What you need to know about RAID

What is RAID? RAID is actually an acronym for "Redundant Array of Inexpensive Disks," meaning that there are two or more disk drive that have been configured (striped) to behave as one drive. In the world of video, this allows for much faster read/write times than can be achieved using a single drive. How? Imagine that each drive has one 'scout' looking for information on that drive: if a drive is acting alone, it only has one 'scout,' however if it has been striped together with another drive (as a RAID array), there are now two 'scouts' out looking for the same information, allowing for the information to be retrieved twice as fast.

By striping more drives together, more speed is achieved. This is especially helpful with multiple streams of HD video, where the rule of thumb is that you should have a minimum of 200 MB per/sec read/write speed per stream to avoid having dropped frames. Typically, eight drives or more are striped together to achieve the desired throughput.

In the mid to late 1990's, nearly all drives were striped together using software (usually built in to the operating system). This 'software strip' is commonly referred to as RAID 0 (zero) and it allows all the drives to behave as one drive, achieving great data throughput. The down side to this type of striping is that if one drive fails, all media is lost. In a typical SCSI drive setup eight 36GB drives might have been striped together, so by having one drive fail you would lose 288GBs of media.

In the days of SCSI drives - the 1990s - their failure rate was quite good - usually one million hours mean time between failures; the downside was that SCSI drives were very expensive. Today, most video drive providers have moved to much cheaper SATA drives with much higher capacities. Unfortunately, SATA drives are not nearly as robust as the 'old' SCSI drives. This creates a potentially devastating problem: large amounts of media that have a greater likelihood for failure. Instead of the 36GB SCSI drives, today you might have eight one terabyte (TB) drives striped together; now when you have a drive failure you will lose all eight terabytes of data.

Many video editors are unaware of the dangers striping large SATA drives together using RAID 0. In today's tapeless workflow, where P2 and similar cards are off-loaded and then re-used, the media on your drives could be the only copy that exits: lose that media, and you have to re-shoot - usually not an option.

Protecting your media using RAID

Thankfully, there are other versions (usually referred to as levels) of RAID besides RAID 0 that can and will help you avert disaster. The key with the other levels of RAID is that they save what's called parity information on the drives. The parity information allows one (sometime two) drives to fail without losing the data (your media).

While there are several levels of RAID that use parity information, there are really only two that are appropriate for video editing - RAID 5 and RAID 6, as they tend to be able to handle the parity information and maintain a high data throughput necessary for video. If you would like to learn all about all levels of RAID, check it out on Wikipedia.

In short, both RAID 5 and RAID 6 put parity information on all drives that have been striped together, the difference being that RAID 5 can only handle one drive failure while RAID 6 can handle two drives failing. The upside to RAID 5 (and a reason that most editing drives are set up as RAID 5) is that the parity information takes up about the amount of space as one of your drives in the array. For example, an 8 drive system with 1TB drives, once striped as RAID 5, would end up with about 7TB of usable storage space. A RAID 6 setup using the same scenario would yield about 6TB of usable drive space.

What's needed for RAID 5 or RAID 6

Both RAID 5 and 6 require a controller to manage the data. The controller can either be in the drive chassis or on a card that goes in the computer. Some of the more sophisticated systems actually have dual controllers for even better performance. In either case, the controller is basically a computer managing the data, and to configure the drives there is either a separate software used or many systems have a network connection and are configured using a web browser. Many of the RAID controllers also offer an incredibly valuable feature: the ability to send out email alerts should a drive (or other component) fail. This can be especially handy for drive systems that are located in a machine room where a warning beep or light flashing may go un-noticed.

The cost of RAID 5 or 6

In the past, the cost for a protected RAID system was usually in excess of $5,000. Today, that cost has dropped significantly, with RAID 5 systems nearing the $1,000 mark. Ultimately, RAID 5 or RAID 6 is cheap insurance against the inevitable drive failure. Post-Op Video offers RAID solutions from Rorke Data, Sonnet Technologies, CalDigit, LaCie, G-Technologies and Editshare.