RAID advice...

[drewson99]

I am thinking of using a RAID - what are the benefits, and which type of RAID would be best for my purposes?

I spend a lot of time backing up DVDs but also capping 12GB DV AVIs and sometimes have several I am working on at a time.

Also, I am wondering - the name "redundant" implies to me that info is written more than once to ensure against errors? does this mean that it takes more than 100GB of RAID space to contain a 100GB file?

should I get a RAID knowing what you know about my system and my activities?

I own: 4 WD HDs, all 7200rpm (28GB, 60GB, 100GB, 120GB)

I'd buy more if it was necessary. also - what other hardware is required for a RAID?

thanks all,

Andy

[ptmurphy]

Well, without going into a detail of RAID, it depends on the level of RAID you implement.

Basically you can do three levels.

Level 0 - basically just makes two physical drives look like one.

Shadow - Exactly as it sounds, two disks contain exactly the same data, so if one fails, the other takes over.

What I call true RAID - Multiple disks (3+) are used to ensure that if any one disk fails all the data is still available. So essentially, all data is written twice.

Hope this helps, as it is a very basic description of RAID.

[SaSi]

To create a RAID array, you should use identical drives (not just the same size - which is crucial), but also same speed and performance. Otherwise the RAID will become inbalanced and perform poorly.

IDE RAID controllers are in all m/b nowadays. They cost next to nothing and perform as much as their cost implies :P

With IDE RAID controllers you can either build striped arrays, where more than one drives are accessed in parallel to enhance performance, or mirrors, to safeguard your data.

I use two of my 200Gb drives as a stripe to maximize throughput while capturing. It works wonders as it gives me a twice as large filesystem and almost twice as much write speed.

For Video editing and stuff like that, I don't care about reliability as the drives are used to store a few huge files for a few days.

If you want "real" RAID, you should go SCSI. Ultra 160 or Ultra320 controllers and disk are available, and in striped mode can offer you at least 2.5 ~ 3 times as much performance as IDE do. However, they tend to cost almost 10 times as much as IDE do.

[sterno]

What I call true RAID - Multiple disks (3+) are used to ensure that if any one disk fails all the data is still available. So essentially, all data is written twice.

This is wrong. Data is written twice when you use RAID1 (what you called shadowing). RAID5 (which is what you're describing here) writes each block of data once, but calculates a block of parity data for every two blocks of data to do error correction. For every two blocks of data written, it puts the blocks on two different physical drives and calculates the error correction data, which gets written to a third drive. If any one drive fails, it uses the error correction and the remaining data stored on the other drives to recreate the data that was on the third drive. It makes more efficient use of your disk space, but controllers that do it well are pretty expensive because they have to calculate the error correction data on each write. Benchmarks have found that a lot of IDE-RAID controllers that offer RAID5 perform writes very slowly because the manufacturers went cheap and didn't put a fast processor on the card.

With IDE RAID controllers you can either build striped arrays, where more than one drives are accessed in parallel to enhance performance, or mirrors, to safeguard your data.

If you buy a decent controller (like a high-end 3ware or Adaptec model) you also get RAID5, which makes more efficient use of your drives than RAID1 but provides almost as much safety. Some controllers also let you combine mirroring and striping (usually called 10 or 0+1, minimum 4 drives) or mirroring and parity (called 50 or 0+5, min. 6 drives). Those are of very limited practical use, though.