Is Raid 0 worth it for a g5 DAW?

[The Sinner]

does this make much difference on a mac?

[Frodo]

Great question, Sinner:

Check out this report, granted-- it's an older report, but it might shed some initial insite:

http://www.barefeats.com/hard35.html

and

http://www.barefeats.com/hard53.html

http://www.barefeats.com/hard40.html


I've been wondering about this myself and am searching for more recent reports on the newer computers.

[giles117]

In DP it makes a noticiable difference in the proc meter (Especially with high track counts)

Mty only issue is the fault tolerance of a RAID array. I have done something as simple as do a software update and lost access to my array. I havnt read too much on that issue.

It has made me a tad gun shy (software raid)

If I try it again, I might do so with the mac pro, or with a hardware card.

[sdemott]

Not to be a total geek here, but RAID 0 is not, by definition, a true RAID array (The "R" in RAID standing for Redundant). RAID 0 is a set of striped drives (no redundancy = no data protection).

Yes - it's fast as all getout as far as read/write access goes, but should anything go wrong with either drive then ALL the data is lost on BOTH drives. For this reason I would suggest a RAID 0+1 or a RAID 1. They are the fastest performers where read/write access is concerned. The main difference being the cost. RAID 1 only requires 2 disks, while RAID 0+1 requires at least 4 disks.

The other RAID types (3,4,5) all suffer from read or write bottlenecks because of the parity data, making them poor choices where large file read/write access is required. And RAID 2 is just not needed with todays drives.

Not sure if that helps or just further confuses the matter.

[The Sinner]

Not to be a total geek here,

you pretty much are if you are a member here. in good company of course.


Perhaps I should reprhase. does anyone sucessfully use raid (any) with DP and see a noticeable difference in tracks and plug in streaming??

[newrigel]

does this make much difference on a mac?

Not really, how you would get a performance bump is to record the audio to a specific drive and then DP will look @ multiple drives this way... I think that the new MP will be so great @ this because of being able to see 4 SATA drives simultaneously so there's where you would reap the benefits...

[Frodo]

If I may ask,

I've got an eSATA II card coming with a 5-drive port multiplier. The card supports 20 HDs (or 4 multipliers with 5 drives each).

I am now concerned as to how much I can reasonably run on this. I understand that with a 5HD RAID, the throughput would be 190 MB/sec+ reads and 230 MB/sec+ writes. That's down a bit from the 3Gb (gigabit) per sec top speeds, but then the PCI host is running on a 133Mb/sec PCI-X slot.

I know all of this is going to give better performance than FW400/800 drives clogged into one bus, but the Mb/sec numbers are not quite adding up. I don't think I'd want to put more than two RAIDs on this one card lest I find myself close to FW800 performance again.

Any thoughts on this?

[HCMarkus]

Any thoughts on this?

You are going to be very happy.

[Frodo]

Any thoughts on this?

You are going to be very happy.

Oh, goodie-goodie-goodie-goodie...

Hey-- I'm now getting mixed messages from different users. Someone from Vienna Symphonic Library customer support recommended NOT using RAID due to added seek times.

It suddenly occurs to me that I know less about RAID configs than I should. I'm now wondering if I shouldn't run these drives as "indies" (in parallel). But until I start building a network of VI farms, I know that with the entire Vienna Symphonic Library loaded into this one array there will still be instance count limits.

Things brings me to my next consideration for how to load these drives. With 5 drives, one would think that strings could go on one, winds on another, brass on another, percussion on another, and so on. But since I will likely have to create tracks by section (strings, brass, winds, etc), it may be better to put, say violins and violas on one drive, and cellos and basses on another drive to increase seek times (?) and ease the strain of streaming should all strings reside on one drive.

So, here's what I'm thinking:

Drive 1
-- violins and violas (large orch)
-- flutes and oboes (picc and Eng horn)
-- trumpet sections

Drive 2
-- cellos and basses (large orch)
-- clarinets and bassoons (bass and contra)
-- trombone sections

Drive 3
-- chamber and solo violins and violas
-- French horn sections
-- Tubas

Drive 4
-- chamber and solo cellos (and basses?)
-- French horn solos
-- other brass solos

Drive 5
-- percussion
-- harp
-- (plus Ivory?)

I know it looks a little discombobulated, but each instrument group is spread across multiple drives.

It's a shame I don't have 5 computers to run these drives independently and concurrently. Until then, I'll have to continue running each section one at a time and then rendering each section out as audio to save resources as more VI sections are loaded.

Okay, so you say- "but Frodo-- VSL is only 550GB. Why do you need 1.25 TB over 5 drives?"

I also have EWQLSO Gold Bundle and have yet to sort out how to deal with that. I may have to go back and spread it out across the five drives the same way. It's only 36GB, so it shouldn't clog things up very much at all. I'll just have to be careful about how many VIs are run at once.

last thoughts--

I'm considering putting my internal 500GB SATA into an eSATA II encloser and running it off the PCI-X host, then getting an internal 10k drive to run things like Ivory and other smaller VIs. The 500GB drive will become my audio or video project drive. My stack of FW drives would then become my archive drives, so I could offload projects from my 500GB SATA onto the FWs as projects are completed.

For BFD and non orchestra VIs, I'm going to have to consider another RAID stack. Hmmm.

It will also be nice to finally wipe a bit of bulk from my system drive. I can hardly wait to get all of this sorted out. It feels like buying extra real estate to adjacent property and building a family room onto the house.

[Frodo]

Okay-- after all of this planning, I've been given a ton of info I have no idea what to do with-- this RAID kit arrives in a few hours and I'm more confused than ever.

Here's what I've been told since my last post:

1. Use Raid 0-- it's faster
2. Don't do RAID 0 because it's prone to data failure
3. Do RAID 5 or higher
4. Forget RAID-- get independent eSATA II enclosures (after spending several hundred $$ on a 5x port multiplier)
5. Don't config as RAID, but use JBOD instead
6. Don't use JBOD config-- too slow
7. Use RAID 0+1
8. Don't use RAID 0+1 because it's more volatile than RAID 10 and cannot survive two HD failures.
9. Don't use RAID 1 because it eats HD space.

The list goes on and on. Much of what I've heard in the past two days has been riddled with terms like "slow transfer rates" and "hard drive failure". Honestly, it has me a little spooked.

I may try RAID 5 or RAID 10, but I'm still wondering if I should try to find 2-3 single enclosures and to put a couple of the HDs on separate parallel eSATA II inputs into the host.

So my question is, what do you recommend for diaper rash?

[solar]

Regarding #4 - Don't know why you would need to get independent eSata II enclosures, aren't you still able to run the drives separately (not in a raid). I thought the mulitplier was just taken advantage of the full bandwidth. You should still be able to run your drives in RAID or Indie, so if you already forked over the cash the enclosure is still useful.

I've thought long and hard and have almost come up with the same conclusion. Indies are probably better for streaming audio and dividing up the work flow among drives is probably most efficient in the realm of data transfer, not the pocket book. Imagine 4-6 raptors streaming.

I actually buy used glyph enclosures on ebay and convert them to SATA enclosures (no one really makes SATA rackmount enclosures yet) it is really easy and you can find hardware to convert them online very easily. I mainly use the 2u racks that are removable. But I have also converted the 1u rackmounts. At the moment I only have the Tempo eSATA 4+4 so I only get 4 ports, but have been thinking of get one of Sonnets port mulitplier cards and using maybe something like this http://www.sataport.com/ in order to split things up outside of the box. Or..... I just bought a Glyph trip recently so I can go ahead and buy a Port multiplier bridgeboard card - http://www.datoptic.com/cgi-bin/web.cgi ... detail=yes - and have at least 4 drives in one unit streaming samples..... Oh and I just thought of this, the 5th port could go back outside of the box daisy chain style into another box...cool (sorry that was for my own personal epiphany):D

At any rate you can get boxes cheap right now and for the diy'er like myself easily convert, even better would be to already have a box and not have to buy one, just the hardware. Either way it is definitely cheaper than the cheapest new FW glyph drives.

I probably have one advantage and that is I have a friend who runs a CNC at his work and cuts out my front panels so I don't have to use the old SCSI glyph panels and only get charged a small fee. But look into it it may be a way to go, and it's rackmountable.

[Frodo]

Regarding #4 - Don't know why you would need to get independent eSata II enclosures, aren't you still able to run the drives separately (not in a raid). I thought the mulitplier was just taken advantage of the full bandwidth. You should still be able to run your drives in RAID or Indie, so if you already forked over the cash the enclosure is still useful.

There was some debate over the JBOD (just a bunch of discs or "concatenated discs") config instead of RAID while still using the 5-disc enclosure. Some say that the computer sees all five discs as one HD, making seek times on a 1.25TB system longer and slower. All data is still flowing down one pipe in that regard. Customer support from one of my VI makers was the one that suggested using Indies would be most ideal-- each drive is accessed independently and concurrently without the higher data failure rate of the RAID config-- 5 data streams at 250 (or less) GB each rather than "the blob" concatentation.

After some time not getting 2 of the 5 drives to mount, I finally managed to get all 5 drives to appear and have only started installing files on them. I've yet to really put the new drives to the test, but will be doing so over the next couple of days.

No, the enclosure will not go to waste. But some setups work better than others on different computers under different circumstances. The only real way to sort that out is by informed trial and error. So far so good, tho.