[ninahagen]

I have found a drive I think will be perfect for storage & backup solutions:

http://www.lacie.com/products/product.htm?pid=10906

I’d like to use one of these for data storage keeping it connected to the Mac Pro 24/7. The other would be kept in a waterproof container in a fireproof safe, excepting a once weekly backup of all data on the storage drive.

I have three questions about this:

1. Can I connect one La Cie to each of the 2 extra sata ports on the motherboard?
(If so, that would free up the 4 bays for a 4 x 150GB Raptor RAID 0 scratch disc.)

2. Am I correct in assuming that the sata ports would provide the fastest connection?
(PCI card and Firewire 800 being the other options)

3. Are there any capacity issues for these ports or OSX? Can they handle 187 mb/s transfer rates?
(spec sheet states that Windows XP, as an example, will not recognize single volumes greater than 2TB)

Have to get this finalized before the imminent release of the next gen MP. Any help would be much appreciated!

[mduell]

1. Yes.
2. Yes.
3. No. Yes.

I'd also look at the WiebeTech RT5e enclosure since it supports RAID5 instead of just 0, 1, 0+1 like the LaCie (RAID5 makes more sense for backup storage, IMO). Would be $2700 for 3.75TB (3TB after RAID5) or $3350 for 5TB (4TB after RAID5).

[ninahagen]

Originally Posted by mduell 

1. Yes.
2. Yes.
3. No. Yes.

I'd also look at the WiebeTech RT5e enclosure since it supports RAID5 instead of just 0, 1, 0+1 like the LaCie (RAID5 makes more sense for backup storage, IMO). Would be $2700 for 3.75TB (3TB after RAID5) or $3350 for 5TB (4TB after RAID5).

Thanks again for your great suggestions!

I’d like to bring up a a few further issues:

1. It is thrilling that one can use such a big RAID volumes, and only use up a single sata port each. (I figured there would be a problem with capacity, Apple RAID card software or something else). The Wiebe site’s RAID faq says “Writing to 10 drives is roughly 10 times faster than writing to 1 drive.” I had no idea speed scaled in such a linear way. Where is the theoretical rolloff? Can you gain 100 times the speed with 100 HDs? Is the limiter the speed of the sata port, RAID software or hardware?

2. Webie also offers a user configurable version. Could one install 5 x Cheetah 15k rpm drives, and would they be faster than 5 Raptors? How could one estimate the mb/s of both? I know that access/read time and latency play a role too.

3. Regarding RAID 5: I wanted to keep the capacities the same on the storage array and the backup array. For the storage array, these would include big, big picture files (1~2G) I use regularly, so it seem the storage array should be RAID 0. Wouldn’t an identical RAID 0 backed up weekly be safe enough? How much of a hit in speed would I take in RAID 5?

4. Is there any sense to RAIDing the system drive?

[OreoCookie]

1. A RAID0 scales pretty much linear in the number of drives for practically relevant applications. A RAID5 is much faster than a single drive reading things, but about as fast as a single drive writing data. RAID5 controllers usually have a sizeable cache these days that speed up things considerably.
2. Yes, they'd be faster.
3. You shouldn't have a RAID0 backup volume, definitely, definitely not. Do not merely clone your volume onto the backup volume. That's disaster meant to happen. Another professional photographer uses separate drives for each job (he simply bills his customers), another uses a large WiebeTech RAID5. Read more here for real-world solutions of real members.

[mduell]

Originally Posted by ninahagen 

1. It is thrilling that one can use such a big RAID volumes, and only use up a single sata port each. (I figured there would be a problem with capacity, Apple RAID card software or something else). The Wiebe site’s RAID faq says “Writing to 10 drives is roughly 10 times faster than writing to 1 drive.” I had no idea speed scaled in such a linear way. Where is the theoretical rolloff? Can you gain 100 times the speed with 100 HDs? Is the limiter the speed of the sata port, RAID software or hardware?

2. Webie also offers a user configurable version. Could one install 5 x Cheetah 15k rpm drives, and would they be faster than 5 Raptors? How could one estimate the mb/s of both? I know that access/read time and latency play a role too.

3. Regarding RAID 5: I wanted to keep the capacities the same on the storage array and the backup array. For the storage array, these would include big, big picture files (1~2G) I use regularly, so it seem the storage array should be RAID 0. Wouldn’t an identical RAID 0 backed up weekly be safe enough? How much of a hit in speed would I take in RAID 5?

4. Is there any sense to RAIDing the system drive?

1. It's pretty much a linear gain in bandwidth (assuming you have a bus on the other side of the controller to support it), but no gain in latency. That's why a lot of high end servers have arrays with 100+ drives. A single SATA port becomes the limiting point after about 4 drives (~280MBps practical for SATA 3Gb/s and 70MBps per drive is typical).

2. No, you can't use either Cheetahs (SAS) or Raptors (SATA) in the RT5; it only supports IDE disks.

3. You're just going to do a straight copy, no versioning or anything? I have no objection to using RAID0 for each array. RAID5 has the same read speed as RAID0, but typically the write speed of only one drive.

4. Not really, and it can cause a lot of hassle for installing firmware updates and such. I'd suggest a Raptor for system, a Raptor for apps, and then 3 for your scratch array.

[ninahagen]

Thanks so much for your illuminating comments.

Originally Posted by OreoCookie 

A RAID0 scales pretty much linear in the number of drives for practically relevant applications.

Originally Posted by mduell 

It's pretty much a linear gain in bandwidth (assuming you have a bus on the other side of the controller to support it), but no gain in latency. That's why a lot of high end servers have arrays with 100+ drives.

1. That is a magnificent fact.

Originally Posted by mduell 

A single SATA port becomes the limiting point after about 4 drives (~280MBps practical for SATA 3Gb/s and 70MBps per drive is typical).

2. Does it matter to the sata port how many drives create the 280 mb/s, or is it only the transfer rate that limits?

3. Also, the RT5e spec sheet gives a transfer rate of 45mb/s. The LaCie is up to 187mb/s. (?) What am I missing?

Originally Posted by OreoCookie 

2. Yes, they'd be faster.

Originally Posted by mduell 

No, you can't use either Cheetahs (SAS) or Raptors (SATA) in the RT5; it only supports IDE disks.

4. It seems that the Cheetahs would be faster if it were possible to use them. Is there any such five disc array enclosure that can use Cheetahs and port them to a simgle sata port? Maybe the LaCie could, or another rand I am not aware of.

5. Failing that, what is the fastest IDE disc on planet earth?

Originally Posted by OreoCookie 

You shouldn't have a RAID0 backup volume, definitely, definitely not.

Originally Posted by mduell 

I have no objection to using RAID0 for each array.

6. Would each of you be willing to explain why? I figure since the backup array will be effectively off-site (waterproof/fireproof safe), they would both have to fail at the same time for me to lose data. Though I just thought that if the RAID controller went "ape-****" during weekly backups, I would then be vulnerable to total loss on both drives, right? Maybe the solution would be RAID 0 for the storage array, and a larger RAID5 array just in case... although wouldn't that have the same RAID controller vulnerability?

Originally Posted by OreoCookie 

Do not merely clone your volume onto the backup volume. That's disaster meant to happen.

7. By clone you mean RAID1 right? I meant wipe and copy onto the backup array every Thurday.

Originally Posted by OreoCookie 

Another professional photographer uses separate drives for each job (he simply bills his customers), another uses a large WiebeTech RAID5. Read more here for real-world solutions of real members.

8. Thanks for that... will do this reading.

Originally Posted by mduell 

Not really, and it can cause a lot of hassle for installing firmware updates and such.

9. OK, all clear.

Originally Posted by mduell 

I'd suggest a Raptor for system, a Raptor for apps, and then 3 for your scratch array.

10. Most of the suggestions I have gotten are for putting the system and apps on the same Raptor. What would the advantage be of separting them? Would it outweigh the advantages of a fourth disc in the scratch array?

11. What is the fastest SAS Cheetah out there, and would that have to be run through a PCIe card? If I mounted that in the second optical bay, and ran it through a PCIe card, would that exceed the peformance of a Raptor ported directly to the motherboard?

Thanks for the education!

[OreoCookie]

3. The spec sheet reads 85 MB/s. You're also missing that you have redundancy, the 187 MB/s are only reached when using a RAID0 (which is doesn't protect you against drive failure).
7. No, by clone I mean Carbon Copy Cloner cloning, not a RAID1. Both are not backup solutions.

Just have a look at the two threads I've linked to, especially if you are going to make an investment like this. Most of the topics (including backups) are covered there.

[ninahagen]

Thanks to all your efforts a good strategy is emerging. Of course, the best processor, core density and video card are a given for this kind of system. Beyond that…

RAM: 16GB (4 x 4GB modules in the upper riser)
System+Apps Drive: Cheetah 15k rpm (in the second optical drive bay)
Scratch Drive: 4 x 150GB Raptors in RAID 0 (external, ported to one extra sata port)
Strorage Drive: 4 x Seagate 1TB drives in RAID 0 (internal… in the mac pro bays)
Backup: Webie RT5e 5TB in RAID 5 (4TB usable, weekly Carbon Copy Clone backups)

Some notes & questions on my assumptions follow. Would you please point out any errant thinking?

— RAM: I want to use this system hardware for at least 5 years. That means I will need enough RAM to be ready for CS3 64-bit when it comes out. I will also be managing tons of RAW files in Aperture, and I expect this to be very RAM intensive. We also may end up editing hi-def video (from our scuba trips), which may require even more, so having the second riser free is a plus.

— System+ Apps Drive: Several issues here. Still not sure I can use the Cheetah 15k SAS drive. Does anyone know for certain? Even if I can it has to route through a PCIe card, right? Which would be faster: Cheetah+PCIe or Raptor+sata port (motherboard)?

— Scratch Drive: What is the best way to connect an 4-Raptor external RAID0 array to a single sata port? There is also the La Cie Biggest Quadra option. I would like to approach the 280 mb/s transfer capacity of the sata ports as closely as I can.

— Storage Drive: Does anybody know how to estimate the transfer rate from 4 Seagate Barracuda ES.2 1TB drives in RAID0 via the internal bays and the Apple RAID card?

— I am so grateful to have read Oreocookie’s great link… definitely has to be RAID5… that way I have 3 fallbacks: the main storage array, the backup array, and the parity stripe. Hard to believe all three could fail at once. Also, the internally padded waterproof container and fireproof safe should add the final touch. We just have to wait for Apple to release!!!!!

I would really appreciate hearing from any of you on the remaining issues.

[mduell]

Originally Posted by ninahagen 

2. Does it matter to the sata port how many drives create the 280 mb/s, or is it only the transfer rate that limits?

No, it doesn't matter how many disks.

Originally Posted by ninahagen 

3. Also, the RT5e spec sheet gives a transfer rate of 45mb/s. The LaCie is up to 187mb/s. (?) What am I missing?

RAID5 is hard and requires a lot of math. Also note it's 85MBps on the spec sheet.

Originally Posted by ninahagen 

4. It seems that the Cheetahs would be faster if it were possible to use them. Is there any such five disc array enclosure that can use Cheetahs and port them to a simgle sata port? Maybe the LaCie could, or another rand I am not aware of.

I'm sure there is; if not SATA, then you could get a fiber channel card. At what point is it good enough? What's your realistic budget here? I mean, why aren't you getting a 15 drive array of 15K SAS drives on fiber channel or Infiniband or something similar? At what point are you buying a refrigerator sized disk array from Hitachi?

Originally Posted by ninahagen 

5. Failing that, what is the fastest IDE disc on planet earth?

Any of the 1TB drives.

Originally Posted by ninahagen 

6. Would each of you be willing to explain why? I figure since the backup array will be effectively off-site (waterproof/fireproof safe), they would both have to fail at the same time for me to lose data. Though I just thought that if the RAID controller went "ape-****" during weekly backups, I would then be vulnerable to total loss on both drives, right? Maybe the solution would be RAID 0 for the storage array, and a larger RAID5 array just in case... although wouldn't that have the same RAID controller vulnerability?

I'm OK with RAID0 when there's another copy of the data sitting around.

Originally Posted by ninahagen 

7. By clone you mean RAID1 right? I meant wipe and copy onto the backup array every Thurday.

RAID1 is a mirror. 'Clone' and 'wipe and copy' are the same thing in my book, and neither is a backup. A backup will let you get a file of a copy the way it was a month ago, among other things.

Originally Posted by ninahagen 

10. Most of the suggestions I have gotten are for putting the system and apps on the same Raptor. What would the advantage be of separting them? Would it outweigh the advantages of a fourth disc in the scratch array?

It's a compromise to boost performance while avoiding putting the system drive on a RAID array.

Originally Posted by ninahagen 

11. What is the fastest SAS Cheetah out there, and would that have to be run through a PCIe card? If I mounted that in the second optical bay, and ran it through a PCIe card, would that exceed the peformance of a Raptor ported directly to the motherboard?

300GB/15K, yes, yes.

Originally Posted by ninahagen 

Thanks to all your efforts a good strategy is emerging. Of course, the best processor, core density and video card are a given for this kind of system. Beyond that…

RAM: 16GB (4 x 4GB modules in the upper riser)

You'd want 2 modules on each riser; more bandwidth, less latency.

Originally Posted by ninahagen 

— RAM: I want to use this system hardware for at least 5 years. That means I will need enough RAM to be ready for CS3 64-bit when it comes out. I will also be managing tons of RAW files in Aperture, and I expect this to be very RAM intensive. We also may end up editing hi-def video (from our scuba trips), which may require even more, so having the second riser free is a plus.

At the performance level you desire, 5 years is frankly unrealistic; some technology will change, creating a silly bottleneck compared to how much you've spent on the system. I think you're better off buying a $15k system every 2 years than a $30k system every 4.

Originally Posted by ninahagen 

— System+ Apps Drive: Several issues here. Still not sure I can use the Cheetah 15k SAS drive. Does anyone know for certain? Even if I can it has to route through a PCIe card, right? Which would be faster: Cheetah+PCIe or Raptor+sata port (motherboard)?

I don't know of any OS X compatible PCIe SAS cards, but I assume one exists. The former is faster.

Originally Posted by ninahagen 

— Scratch Drive: What is the best way to connect an 4-Raptor external RAID0 array to a single sata port? There is also the La Cie Biggest Quadra option. I would like to approach the 280 mb/s transfer capacity of the sata ports as closely as I can.

I really don't think you should be putting your scratch drive externally.

Originally Posted by ninahagen 

— Storage Drive: Does anybody know how to estimate the transfer rate from 4 Seagate Barracuda ES.2 1TB drives in RAID0 via the internal bays and the Apple RAID card?

About 200-300MBps; 400MBps is a good estimate for the upper bound, and perhaps 100-150MBps as a lower bound.

[mduell]

While it's amusing to randomly throw out the fastest product you've seen on a web page somewhere, I don't know that we're really helping guide you toward the best system for you.

My first question is what is your budget? I don't care how you specify it (a range or a maximum or whatever), but there are a lot of tradeoffs to be made creating a balanced system within a realistic budget.

Second question is what do you already own? Reuse is always good when practical.

Third, what's the rest of your setup like? Do you need to buy displays inside your budget? How is your network setup?

Fourth, have you done a quantitative analysis of your needs? How much data do you have, how fast is it growing, and what's it worth? What size files and media streams are you working with?

It may be that a few Raptors and 1TB drives is enough to cover you for a couple years; or it may be that you really need a 15 drive SAS array on some exotic interconnect.

[ninahagen]

Originally Posted by mduell 

While it's amusing to randomly throw out the fastest product you've seen on a web page somewhere, I don't know that we're really helping guide you toward the best system for you.

Well, I am learning lots thanks to you and I agree on the need for balance.

Originally Posted by mduell 

My first question is what is your budget? I don't care how you specify it (a range or a maximum or whatever), but there are a lot of tradeoffs to be made creating a balanced system within a realistic budget.

About 2 million yen... $17~18,000.

Originally Posted by mduell 

Second question is what do you already own? Reuse is always good when practical.

A G5 Quad 2.5 with 8GB RAM. (and 3 lesser macs)

Originally Posted by mduell 

Third, what's the rest of your setup like? Do you need to buy displays inside your budget? How is your network setup?

Just a single 30" display. No network. My staff and I swap files off memory sticks, the office's 80GB iPod, CDs, and DVDs.

Originally Posted by mduell 

Fourth, have you done a quantitative analysis of your needs? How much data do you have, how fast is it growing, and what's it worth? What size files and media streams are you working with?

Not an extensive analysis... just simple... we have about 1TB of data, and it grows in spurts, depending on the jobs we have. Since we will start editing hi-def scuba diving videos (a hobby). What is is worth? Well, we are preparing an anthology of paintings by a 19th Buddhist nun... the works come from collections all over to be scanned and photographed, and we have realistically one chance. We also have nearly 900 translations of her poems that will be wedded with hi-res images. Plus countless other images of artwork we have handled over the years, important to guiding new clients. Losing our data would be a disaster, not one that could not be undone, but a serious drag. File sizes are often 1~2GB. The G5 handles those fine, but batching in CS where the things can really get time consuming.

Originally Posted by mduell 

It may be that a few Raptors and 1TB drives is enough to cover you for a couple years; or it may be that you really need a 15 drive SAS array on some exotic interconnect.

I must admit, I do enjoy buying the best at the time... that was the rationale for the G5 Quad... it has served us in ways we could not have imagined... and I am willing to set a medium high budget and get the best I can within it. I thought this might be it:

5500 Mac Pro (hi-end vid card, RAID card)
2650 OWC RAM (4 x 4GB)
3350 Webie RT5e Backup (5TB in RAID5=4TB)
1100 Raptors for RAID0 Scratch Drive (4 x 150GB)
300 External Enclosure for Scratch Array
400 Cheetah + PCIe card (system & applications)
300 Labor & Bracket for Cheetah
1400 Hitachi Deskstars for Storage Drive (4 x 1TB)
2000 Monitor

17000 Total

Given the uses I outlined, how would you spend that money, plus or minus $1000?

[UnixMac]

Get the fastest machine you can afford.. for me that was a similar albeit less RAM, and fewer Raptors, cheaper RAID0 boxes (I have two identical so I can back one up on the other)..

I had the fun/pleasure of helping my brother-in-law build a machine for his home studio which he uses to make both music and video with.. he had the $ for 16GB RAM, and an X-RAID at 7TB and cache module, and spare parts. Add to that 2x30" displays and a high end APC backup.. I was quite jealous needless to say, but the reality is that he had the commercial need for such a system and his producer liked only working with Macs.. so it was an easy job.

Simple. get the best you can afford (period)... afford being the operative word (which is to say pay for with cash!).

[OreoCookie]

I wouldn't get 16 GB RAM right away: most of the apps cannot use more than 32 bit, so get 6, perhaps 8 and the see whether you actually use all of it (the relevant number here is the number of page-outs and not free RAM!). If you don't see any page-outs (a few hundred is ok, more than a few thousand and you need more RAM), then don't buy more RAM. RAM gets cheaper with time as do all components.