Question about RAID1

IsaacKuo

RAID1 doesn't work like that. All it does is mirror your data; as far as your OS-apps are concerned, they're talking to one single volume. The RAID control mechanism (be it hard-hard, soft-hard or soft RAID) synchronises the primary volume with the mirror. This is, to intents and purposes, invisible to the user. No advantage in throughput, this is the simplest data redundancy configuration (one half of the mirror set can fail, the other set takes over as primary, replacement drives are automatically synchronised - which may hit performance).

RAID0 (striping) works by striping data at the block level; this is not a true RAID and should not be used in mission-critical environments because there is NO FAULT TOLERANCE. Recommended for high-throughput processing such as video-image processing (where the advantage of this method shows).

On your primary qwuestion of overall system performance, my suggestion is to parbreastion your volume setup so that you can take advantage of the different setups. One such (extreme) setup is detailed below.

SYSTEM: flat, one drive or RAID1 - fault tolerance isn't absolutely necessary here, tho it may help if you don't have the channel space for RAID to burn a ghost image to DVD of your setup once you have it the way you want it - makes recovery that much less painful. Minimum 2 disks.

USER TEMPORARY FILES: RAID0 - this is your "scratch" drive. Use a couple old drives for this - 4GB each should be plenty (go 8GB each if you're encoding dual-layer DVD), for up to 16GB of very high throughput processing space. Minimum 2 disks.

USER DATA: RAID6 - of course, nothing beats DVD and a fireproof safe, offsite and stored in a helium environment. RAID6 uses dual-parity distributed across the array, hence offers extremely high fault tolerance as well as increased performance over a single drive or a RAID5 array (where single-parity distributed across the drives offers less overhead=more efficient use of space but less fault tolerance). Minimum of four disks.

OK, not so extreme - but the idea here is to have a pretty bulletproof system while reducing the bottleneck on the IDE bus by maximising the use of the channels available.

NOTE: SYSTEM and USER TEMPORARY FILES can use the same pair of disks - say, a pair of 20's, split 12-8 respectively, then the 12GB parbreastions dedicated to SYSTEM in a RAID1 configuration and the 8GB parbreastions dedicated to USER TEMPORARY FILES in RAID0. Using the above configuration, this takes the drive overhead from 8 units to 6.

SPECIAL NOTE: for any RAID array, it is recommended to use the same make and model drives to build each array. Performance will be hit (if the array works at all) if you use two different sized drives to build an array. On the same note, an 80GB drive from Hitachi will NOT be the same size as an 80GB drive from Maxtor (in fact there's an almost 3GB real difference between those two - purely a result of the specifications using different scales (tho some of it is down to the filesystem, controller and-or parbreastion manager being used)).

Also note that RAID arrays should be built across channels (ie one drive in the array per IDE channel) to maximise performance; this is to say, an 8-disk array, using A as SYSTEM(2), B as TEMP(2), and C as DATA(4) is configured thusly:

Controller channel* Array 1 A, C 2 A, C 3 B, C 4 B, C

*this would buttume, of course, that you have 4 IDE (PATA) channels available to you (8 ports) and are using some level of soft-hard RAID setup (not necessarily but it helps when setting up the SYSTEM and TEMP arrays). Hard-hard PATA RAID controllers are obscenely expensive, soft-hard are cheap (like, GBP£20 a pop). For SATA you would need 8 SATA channels (almost certainly requiring an offboard controller).

-- When all else fails... Use a hammer.

Some people are like Slinkies They serve no particular purpose But they bring a smile to your face When you push them down the stairs.