Welcome to the MacNN Forums.

Hi:
I have just ordered a 2008 2.8 Mac Pro with 320gb hard drive. It is supposed to be delivered the end of the month. I was planning on adding a 500gb or possibly 750 gb hard drive. I was considering the Seagate 7200.11 SATA II 32 mb cache hard drive but I am having second thoughts b/c of several postings on this and other websites noting problems with the Seagate hard drives and Mac Pro. Are these isolated problems or should I consider other hard drives?
I also ordered the standard 2 gb of memory and will be adding an additional 8gb. Will be using Aperture, Photoshop CS3 etc.
Plan on getting one or more additional hard drives at 500 or 750 gb with 32 mb of cache. (Retail versions since some posters have recommended that oem versions should be avoided.)
The Hitachi enterprise hard drive has 32 mb of cache but it is $50 to $75 more. Any suggestions

Cache doesn't really matter; yea, it's nice if you're doing a lot of small writes but not a huge deal for general usage. The Seagate/Mac Pro issue was specific to one firmware revision (3.AEE IIRC) for the 7200.10.

With a machine in your price/performance range, I'd go with a 150GB Raptor for scratch and 1TB Seagate or Hitachi for storage. Keep in mind you want to keep your working drive about half empty for performance, so 1TB gives you 500GB usable. The Hitachi has the edge for I/O performance in 1TB drives; I wouldn't bother paying for the Ultrastar enterprise drive. There's nothing wrong with OEM drives, they carry the same 3 or 5 year (depends on brand) warranty as their retail counterparts.

150GB Raptor for $170
1TB Hitachi for $272

1. What, if anything, should I do with the 320 gb hard drive
that will come with the standard Mac Pro configuration?
Should I put it in one of the other hard drive slots and put an extra copy of my raw and jpeg photos on it.
2. I am planning on adding a Sonnet Tempo SATA E2P
PCI Express card with two external esata connections.
I was going to connect an external drive for time machine via esata -- if I put a 1 T hard drive on the Mac Pro, how large should the external drive be? -- also I was thinking of LaCie unless someone recommends an alternative.

I am planning on adding Fusion, Windows XP and Wordperfect. I will also be accessing the office through a VPN through Fusion/Windows XP.
Should I put them on the Raptor drive?

Use it for OSX/Apps/Windows.

No need to move it. Why make copies of your photos?

Bad idea, doesn't work with 10.5 on Intel. Either wait for new drivers or pay up for the $250 E4P or just buy a $5 bracket to connect the extra SATA connectors to eSATA ports.

1TB should be fine for your TM volume; going above that becomes kinda painful. I'd stay away from LaCie (poor experience and poor prices). Buy another 1TB Hitachi and an external enclosure with the ports you desire (USB, FW, eSATA).

Put them on the 320GB. The Raptor is for scratch.

Great post, mduell.

IMO you will want more capacity than you seem to be going toward if you are using Aperture and CS3 like I am. Note that hard drives start slowing at something like 50% full and can get very slow or even unstable above 85% full; a good rule of thumb is not to exceed 70% full. Raptors are fast but very low capacity. It seems me that for our needs a 2 or 3-drive RAID0 array of 1 TB sized drives is a better way to proceed. Note too that our apps do lots of small reads and writes; anecdotal reports suggest that 32 MB drive buffers may help significantly.

I will probably remove the stock 320 GB drive and put it in an enclosure for external usage. Currently I am debating (all 1 TB sized drives) among the below options. Comments/critiques Option 1 vs Option 2 are much appreciated.
---------------------
Option 1
Drive A system and apps
Drives B,C RAID0 array for Aperture Library, images and scratch
Drive D onsite backup
---------------------
Option 2
Drives A,B,C RAID0 array for system, apps, Aperture Library, images and scratch
System and apps on their own partition to facilitate cloning using SuperDuper
Drive D onsite backup
--------------------
Variation A
Drive E in the extra optical bay
Drives D,E configured as a RAID1 array; when D/E fill, reconfigure as RAID0.
--------------------
Variation B
Drive E in the extra optical bay
Drives D,E configured as a RAID0 array with hourly backups from the B/C or A/B/C RAID0.
--------------------
Variation C1
Drive E in the extra optical bay
Option 1 with RAID 10 configuration
--------------------
Variation C2
Drive E in the extra optical bay
Option 2 with RAID 10 configuration
--------------------

RAID0 array drives (unless being used for backup) will not be allowed to exceed 50% full (note the impact on actual usable capacity). OS 10.5.2.

I wrote down my thoughts on drives for my own reasons, just to organize my thinking, shared below, long:
================================================== ===========
Backup and how we configure it is actually key to the hard drive configuration process. IMO we should first define our specific backup protocol and drives. My analysis follows, an exercise of brainstorming my individual needs. Please feel free to critique/comment:

First question is what can one afford to lose? E.g. in financial transactions, nothing, real time (hence mirror backup strategies). In photography not one image. However we do not care about real-time because images are static chunks of data. IMO images need to be backed up asap after capture such that at least 2 copies always exist. In my case often right on site I copy CF originals to MBP hard drive and burn a DVD before reformatting CF.

Once originals are backed up all that can get lost is edits/organizational work. How much loss can one tolerate? A minute? An hour? A week? Before we can answer we need a reasonable expectation of how often we can expect the catastrophes we are backing up against: fire (been there, East Bay Hills Fire), earthquake (been there), theft (ditto), drive failure (guaranteed to happen at some point).

If the cumulative average time of failure was every 5 minutes even backing up every minute might not be frequent enough. Fortunately with modern drives we can expect such events to average every few years rather than months, days or weeks. And often hard drives presage failure by conveniently making inappropriate noises, giving us time to backup and replace.

On Site Backup
IMO for planning purposes on a typical 3-drive RAID0 array single-workstation setup we can estimate a catastrophic hard drive failure every 1-2 years. So assuming every 1-2 years we lose our RAID0 array, how much editing/organizing time are we willing to redo every 1-2 years? Each person makes her/his own choice; for me I would say 1-3 days. Ergo daily backup on site. Note that 2 or 5 drives of RAID0 changes the probability of failure, but for me and my personal risk acceptance level I would still end up with daily onsite backup.

Daily backup need not be fast, so a single large hard drive should suffice.

The system/apps drive deserves separate consideration because on the one hand everything there is available elsewhere, no irreplaceable data to lose, but OTOH reinstalling all apps and the OS is an all day task. Fortunately we seldom change apps/OS, so backup protocol is simple: clone the drive to off site concurrent with regular off site backup protocol.

Off Site Backup
Certain events - primarily fire and theft - may take out all the drives and the computer as well. These are true catastrophes and very individualized. At one time I had offices in a bad neighborhood where the probability of hardware theft on a monthly basis approached 100%, so hard drives were stored in vaults at night. Today my offices are very low risk and alarmed, but in a 120-year-old historic building, so fire is the major risk. We keep our building clean and well managed and a fire station manned 24/7 is 100 meters away, so I put our full-loss fire risk as very low. Each user has different risk, and how quickly system reinstatement must occur post disaster is also very relevant.

In my personal example a fire would wreck h*ll with my client's operations from an accounting and stores management standpoint because of lost paper. If I could bring the Mac and PC computers back in 2 weeks I would be perceived as a hero, because the paper would never come back. New hardware can easily be purchased and reinstalled within a week at one of our other locations and it would take at least a week to establish the other location.

What that says for my world is that consequent to fire/theft I can easily tolerate 2-3 weeks or more loss of editing/organizing time. Ergo plan on routine weekly backup to off site (our accounting data is backed up off site daily via an employee taking a CD home).

Off site backup done only once a week need not be particularly fast. A TB of data can be stored in a transportable external FW 800 drive. I will just buy a 1 TB external drive and add another when needed.

-Allen Wicks

A Raptor dedicated to scratch is good if one has plenty of drive slots available and is only dealing with Photoshop. However Aperture's Library will require throughput speed plus capacity, and if you use a drive slot just for scratch that limits your RAID options because the more drives in a RAID array the better the array performs.

-Allen Wicks

Here is a typical hard drive performance curve (in this case, a Hitachi 1TB 7K1000).

Once you get to 20-25% full, you've blown your access time by a factor of two and you're halfway to the full disk access time. The transfer rate degradation also starts at about 20-25%, but isn't too bad until out around 70-75%. I would not expect instability or unreliability at any capacity level. My rule of thumb is to buy storage for 25-35% initial capacity and replace at 75% capacity.

Disk cache works well for small writes (since it doesn't care where they go), but does nothing for reads (since it can't predict what will be read). If you're routinely working with files where 16 vs 32MB makes a difference (which is right around where digital photography is today), there may be some tangible benefit to the larger caches. Good news is that all 1TB drives (from major brands) are 32MB cache so there isn't really any choice.

My suggestion is a 150G Raptor in the optical bay and 4x1TB in RAID0 or RAID10. The Raptor for the OS/apps/housekeeping files (like email which are mirrored on a server somewhere) and I'd be surprised if those exceeded 50GB today. The tradeoff between RAID0 and RAID10 is really a tradeoff between how soon you can be back working after a drive failure (1 day with RAID0 if you have a cold spare, 3 if you don't, and no time with RAID10) against how soon you'd have to replace the drives due to storage growth. The performance, particularly for reads, should be largely the same with modern RAID controllers employing striped reads on mirror sets. If your storage needs today are 400-650GB I'd go with RAID10 and RAID0 if larger; this also depends on growth rate, with a high growth rate favoring RAID0 even with modest current needs.

Good advice, follow it. No other comments on your backup strategy.

The improvement in stripe performance, particularly with software RAID, is really pretty marginal after the third disk.

Excellent, Mark. Thank you very much!

I have frequently observed instability as drives get full. Not instability in the drives, but instability in apps as the drive performance deteriorates. However modern drives/apps may be better, I don't know. In any event your 25/35% - 75% rule makes sense.

I will do as you suggest, Raptor for system/apps. For convenience a 4-drive RAID 10 array would be nice. However my growth rate in mass storage needs is high and is likely to get higher when I have a more competent mass storage setup attached to a fast tower. After your counsel I am tending toward these two choices:

---------------------
Option X
Optical Drive: Raptor, system and apps
Drives A, B, C, D: RAID 10 array of 1 TB drives
---------------------
Option Y
Optical Drive: Raptor, system and apps
Drives A, B, C: RAID 0 array of 1 TB drives
Drive D: onsite backup, 1.5 TB or 2 TB size
--------------------

Option X is easy and clean, yields 1.4 TB usable (70%) working capacity mated to 1.4 TB backup capacity (and gives the unnecessary but very significant benefit of real-time backup).

Option Y is more complex and costs more but yields 2.1 TB usable working capacity with to 1.35 or 1.8 TB usable (90%) backup capacity. I apply a 90% full usable capacity when using a single non-RAID backup drive because no apps will ever access data from that drive, speed is not relevant and routine drive maintenance including full reformatting is easy if necessary.

A benefit of Option Y is that by the time I push the capacity of the RAID0 array to its 2.25 TB maximum larger drives will be available and I can simply replace Drive D with a larger drive.

Yet another question: Why is rebuilding a RAID0 more difficult? I would think (assuming a cold spare) that it would be as simple as replacing the drive, reformatting the RAID0 and overnight copying data from the backup. However I have never rebuilt any level of RAID array, so any info is appreciated.

Thanks again. Multiple thanks in fact. After this discussion I will be quite comfortable with either Option X or Option Y.

-Allen Wicks

The problem with Option Y is that you can't get a 1.5 or 2TB drive yet; I haven't even seen them announced (but are expected later this year). Since you now have to go external for backup (I like the SiI Steel Vine based enclosures for this), you may as well build a 4 drive RAID0 internally.

Rebuilding a RAID0 array isn't terribly difficult or time consuming, you're just offline while you do it. With RAID10 you can rebuild while still using the degraded array.

A 1 TB drive will suffice for backup for 6 months or so, perhaps longer because much of my existing data will not be migrated to the new drives. I really would like to have onsite backup internal. Right now I have a half a dozen external drives and I would like to eliminate that with the new setup except for offsite backup, which by definition lives off site.

-Allen Wicks

Then 3+1 sounds like the way to go.

Someone just announced a 1.6TB drive today... but it's flash and U320 SCSI... and probably costs more than my condo.

yea, i'm sure that the SSD 1.6tb drive is going to be OUTRAGEOUS!!!! Considering a 64gb is $500. Not worth it. I've always had great luck with Seagate drives, but the Mac Pro comes with WesterDigital and you can get a WD 1tb for $249 on newegg.

But I would spend the money on a hitachi or seagate personally.

I've been trying to read all the threads that discuss RAIDs in the new Mac Pros, since I plan to order one in a month. I've never used a RAID, but it sounds like it's about time I do. So, in response to expert Sierra Dragon's idea for the RAID in his new Mac Pro I saw:
So this is three drives at RAID 0 plus a fourth drive for daily on-site back-ups, correct? So boot drive, apps, files, scratch are all part of this three drive RAID, and the set-up seems to be approved by both Sierra Dragon and mduell (as appropriate for Sierra Dragon's needs). But then I read this on another thread from another expert:
So my simpleton head spins when it seems you smart people have different ideas of what RAID set-up is appropriate. I would think, in this case, the issue isn't as bad as ninahagen says, because if your RAID went down, it seems like you could just boot from the back-up drive, and the only problem would be the loss of data since the last back-up... Yes? No?

I also read this in response to a post I wrote, thinking RAID 0+1 was the way I would go:
But then I found some charts that seemed to indicate 0+1 has more of a performance boost than Tesselator says:
ImageShack - Hosting :: sisandrasequentialread4yi.jpg
ImageShack - Hosting :: sisandrasequentialwrite9nb.jpg
As well as another page (with links to charts I don't fully understand) that seem to indicate 0+1 is pretty fast compared to a single drive:
Overclock.net - Overclocking.net - View Single Post - * OFFICIAL * HD Database / Benchmark | RAID & non-RAID Drives
The bottom line is I'm trying to understand the speed differences between the different RAID possibilities, and these links seemed to answer that question, unless there are "real world" conditions that these tests don't address (and perhaps that's what Tesselator was referring to with his slower rating of 0+1). Any of you RAID experts want to weigh in? Is there another link I should look at to understand the speeds of RAID set-ups?

Ugh, and I still need a better understanding of seek times vs. data transfer rates and how that effects the myriad of things I do with my Mac (from manipulating huge PSD files to creating music to editing video). My gut is faster data transfer rates will benefit me more than faster seek times, but I don't really know...

What you missed was that Allen would also have a Raptor in the second optical bay as his OS/Apps/housekeeping drive. In Allen's case, the backup is not necessarily intended to be bootable.

I also disagree with Tesselator's RAID10 performance assertions; 4 drive RAID10 performance should be about the same 2 drive RAID0 performance for writes and possibly as good as 4 drive RAID0 performance for reads (if your controller will interleave reads).

Ideally we'd all have giant RAID10 arrays of 300GB 15kRPM SAS drives. Coming down from that to reality, you either have to go with arrays of 150GB 10kRPM disks or 1TB 7.2kRPM disks depending on your storage needs and financial resources. As I said in another thread, RAID10 vs RAID0 is largely about drive failure recovery downtime vs how soon you need to upgrade your disks.

Thanks for clearing that up about Raptor drive in Sierra Dragon's scenario, not sure how I missed that!

Should I be confused by me saying "RAID 0+1" in my examples, and you saying "RAID 10" like it's the same thing?

Again, thanks to you everyone for helping me learn this stuff. I feel like I learn more in this forum than I have reading a lot of online articles.

I just found two decent links on 0+1 vs 1+0:
Nested RAID levels - Wikipedia, the free encyclopedia
Why is RAID 1+0 better than RAID 0+1?