I want to try a raid0 array made of 2 25 GiG partitions only, not whole drives. I have a recent 500 GIG Seagate (5 yr warranty) and 100ATA IDE drive and a 2001 dated 10k Ultra160 IBM (Ultrastar) 36 GIG SCSI.

I read that any raid array causes increased heat to the drives and attached cables etc..., and the SCSI model tends to be hotter than some other drives of that era. I need to boost speed for video editing and will create a total of 50 GIGS and mount my /temp folder there (I will be keeping the edit projects very simple).

Hoping that if a drive fails it's the SCSI first, will I have a problem accessing the non raid part of my IDE drive, considering that my 'targeted for raid' partition is within an extended partition?
Is the ratio of increase in drive wear great compared to gains in I/O and speed increase for this type of work? I found so far yeses and nos, I can't decide.

Hoping to finally re-install hardy studio LTS soon, thnks in adv!

Note: I have an HP workstation of same age as SCSI, with many fans so this means that I have adequate cooling environment, right?

If you are using a hardware raid controller, I don't think you'll be able to have part of the drive in a raid and part not. I have no idea whether software raid will let you do that.

I also don't understand why you think/hope the IBM drive would die first. I have several of those drives, and they are quite sturdy and reliable. I run them 24/7/365 since 2001 and have had exactly one failure - and that one failed a year after I dropped it on a concrete floor. My ultrastars have fans blowing across them and they all run between 35 and 40C (24C ambient), which isn't any hotter than most other drives I mess with, and is noticeably cooler than the Seagate 500 Gig IDE that I have mounted in an external USB case.

I also would not know why you would think that raid puts additional stress on a drive; the stress on the drive is dictated solely by how the drive is used.

When you say HP Workstation do you mean it was originally sold as an PA-RISC HP-UX box (tend to be pretty good; my somewhat dated experience with them suggests they tend to go on and on, with only maybe the hard disks being a bit suspect, but I suspect that is more of a 'bad date range' issue) or a HP high-end commodity PC (better than most, I would suggest, but don't look as if they are engineered to the same standards as the the PA RISC boxes (how they look isn't really relevant, though; it is about how they perform in the long term)- remember they are designed by different divisions of HP.

In any case, they were likely to have adequate, or better, cooling as they came from the factory; what upgrades have done to this is another question.

Well, I can't see that stress on the cables is a factor, unless you mean that you have routed them badly, so the cables themselves are under tension/stressed by vibration.

Of course, anything that multiplies the number of drives multiplies the possibilities for failure, so that is an issue.

This is 3 drives, right?
500 G (interface unstated - SATA, presumably)
100ATA IDE (capacity unstated)
36 G SCSI (Which SCSI variant)

(or are the first two a single drive?)

Your need for speed is based on video; if you are primarily pulling and and writing out large files, I'd just try the most recent drive for your video files; over a few years the tighter data packing tends to overtake the higher rotational speed of the older (at the time) higher end drives.

In this scenario, putting your OS and swap on the SCSI drives would make sense.

If, on the other hand, you are not primarily operating on large files, data density won't help as much.

I'm not sure that there is a good simple answer for this; anything that uses parity, will increase the amount of writes for a particular data block size, but raid 0 doesn't do that. It does distribute the work across drives, which could help if a particular drive is near to a critical temperature; it probably does increase the load on the power supply a little, which would be bad if the power supply is close to critical.

Note also that among the people who have actually measured RAID performance often find their performance isn't as good as they expected; I suspect one of the issues is whether, say, the PCI bridge/interface bus is overloaded. Without knowing more about your arch I can't even guess whether this would apply to you (even if i did know more about your arch I would only be able to guess; not very useful).

Hi! Thanks for the replies,



I have a HP workstation x4000 with dual xeon 2.4 GHz. It has Rambus PC800 ECC RAM not risk chips, but it was an enterprise and not consumer box, used as autocad workstation or server, so it was well built, I beleive.


I have 3 drives but the:

is my description of same 500 GIG IDE drive not SATA.

I have 36 G Fujitsu Man3367MP SCSI (ULTRA160,10k) which supposedly got good ratings in it's day. Windows on it and 11 GIGs left free (for /).

The other SCSI is 36DDYS-T36950 ultrastar by IBM: ULTRA160,10k also but some higher temp. ratings and slighty slower access ratings...

I realise that not having that much video editing knowledge (the files are relatively large but the I/O will be often as well-according to the info. I read on the subject) and no desire to study my PCI bridge interface- and how good it "raits", I might just have slow results one way or another! I still want to try the raid after all-just hope it won't be a headache!

I feel reassured about the hardiness and heat maintenance of my drives and system from what you both said, but jiml8,

I will be using Ubuntu's software raid intergrated into it's kernel, if I understood it correct. see:
https://help.ubuntu.com/8.04/install...e-details.html
and also
According to the raid how-to:
http://tldp.org/HOWTO/Software-RAID-HOWTO.html
It is possible to create raid0 array from 2 partitions!