[SierraDragon]

Originally Posted by misterdna 

...My only question is why not to boot from a RAID.
... a 4-drive single-volume RAID 5 array would be a very straight-forward set-up in the new Mac Pro, and I've read posts in various forums from people who are running this configuration. It seems like I could mirror the single-volume RAID regularly to an external drive (actually two externals, which would be rotated). It seems so simple...

But did you write it down with exact drives and usable capacities? A 4-drive array of any substantial capacity pretty much needs to be backed up to at least a 3-drive array; your "two externals, which would be rotated" turns out to be 2x3=6 drives. And system partitions typically need to be cloned, not just "backed up." You are now talking about lots of drives and complexity.

...I realize there are inherent complications and risks associated with RAIDs. I got the feeling from a recent post of yours that the OS is complicated and RAIDs are complicated, and its just asking for trouble to mix the two. Is it that simple? Is there evidence that this causes problems?...

Sorry, but

...It seems so simple...

and

...I realize there are inherent complications and risks associated with RAIDs.

are absolute contradictions in terms. Along with the need for multiple external drives, inherent complications and RISK are why I choose not to RAID my System drive. Once one writes it down and works out exact backup and recovery protocols the (IMO undesirable) complexity should be apparent.

I have said all along that given enough drives all kinds of RAID arrays including some that include the OS may make sense. Complexity is fully acceptable if enough drives are thrown at the data security issue. Personally I have high capacity needs and (except for offsite backup) I have no intention of throwing more drives at the problem than will fit in my MP; ergo for simplicity the System and apps will not be on a RAID array.

-Allen Wicks

[misterdna]

Originally Posted by mduell 

I'm not trying to be obtuse, I thought the risks of RAID were well known. Wikipedia has reasonable coverage of them.
The annoyance of losing the ability to boot your computer (for a time) because one or four drives failed is quite high in my book. Also should some part of the machine other than the RAID controller or a disk fail, you can't necessarily pull out the drives and stick them in another machine.

But if I regularly cloned a single-volume four-disk RAID 5 array to an external drive, seems like I could just boot from the external and be on my merry way without much down time (other than dealing with whatever was lost since the last back-up). And as time permits, I'd stabilize the RAID and clone to the RAID from the external. Am I being overly simplistic with this view?

Thanks for the link, I had not seen that page. I didn't find anything that addresses specific risks incurred in a system that boots from a RAID.

[misterdna]

Originally Posted by SierraDragon 

But did you write it down with exact drives and usable capacities? A 4-drive array of any substantial capacity pretty much needs to be backed up to at least a 3-drive array; your "two externals, which would be rotated" turns out to be 2x3=6 drives. And system partitions typically need to be cloned, not just "backed up." You are now talking about lots of drives and complexity.

Sorry, but and are absolute contradictions in terms. Along with the need for multiple external drives, inherent complications and RISK are why I choose not to RAID my System drive. Once one writes it down and works out exact backup and recovery protocols the (IMO undesirable) complexity should be apparent.

I have said all along that given enough drives all kinds of RAID arrays including some that include the OS may make sense. Complexity is fully acceptable if enough drives are thrown at the data security issue. Personally I have high capacity needs and (except for offsite backup) I have no intention of throwing more drives at the problem than will fit in my MP; ergo for simplicity the System and apps will not be on a RAID array.

-Allen Wicks

I see simplicity in a computer with a single volume. It's simple to define and interact with, even if it's complicated "under the hood" (such as with a four-drive RAID 5 scenario). Can you see why I see simplicity in that?

Speaking of simplicity, why can't I just clone my four-drive RAID 5 to an external 1TB drive nightly, using SuperDuper!'s smart-update? (actually, I'll rotate several externals). That seems very simple to me, no more complex than the non-RAID backup strategy I currently use (without a RAID). Note that the four drives are for speed rather than capacity, and I don't foresee needing a bigger external back-up (CLONE) drive in the foreseeable future.

If one of my RAID 5 drives goes down, no data is lost, as the three remaining drives contain all my data/apps/sys. If something uglier happens, I boot from my recently cloned external and I am good to go. Seems pretty simple to me.

Please tell me if you think I'm making a major error with my thinking.

Thanks,
Rob

[SierraDragon]

Originally Posted by misterdna 

But if I regularly cloned a single-volume four-disk RAID 5 array to an external drive...

Only be getting to operational detail can drive usage be rationally evaluated ("the devil is in the details..."). Please, write it down. For each drive: what nominal capacity and usable capacity, and the function of each of the 5 drives and whether internal or external and how connected. How and how often backup is to be done (e.g. Time Machine, Super Duper clone, Retrospect, whatever). Software or hardware based RAID control, and what brand.

-Allen Wicks

[misterdna]

Originally Posted by SierraDragon 

Only be getting to operational detail can drive usage be rationally evaluated ("the devil is in the details..."). Please, write it down. For each drive: what nominal capacity and usable capacity, and the function of each of the 5 drives and whether internal or external and how connected. How and how often backup is to be done (e.g. Time Machine, Super Duper clone, Retrospect, whatever). Software or hardware based RAID control, and what brand.

-Allen Wicks

Are you asking me to write it down on a piece of paper to be sure I have done appropriate planning before committing myself down a particular path?

Or are you asking me to post all these details on this thread so you can take the time to personally evaluate the system? If that's the case, I might as while write down all the systems I'm leaning towards, give you all the details of what I think I will be doing with the system, etc. etc. etc. But I doubt that will happen today, I've been here too much today, and am now behind on my work!

[SierraDragon]

Originally Posted by misterdna 

Are you asking me to write it down on a piece of paper to be sure I have done appropriate planning before committing myself down a particular path? Or are you asking me to post all these details on this thread so you can take the time to personally evaluate the system?

I just think your idea of what is simple may change when you get right down to it. Or maybe it really will be simple and some of us here will choose to emulate your proposed solution. Me, I am an anti-paper person. If you lay it out in a text file you can then copy and post it here for discussion. IMO going to the effort of writing all the details down for others to see is helpful - sort of online brainstorming.

-Allen Wicks

[OreoCookie]

Allen is right-on: without giving us specifics of what you need in terms of capacity, it's impossible to tell you whether a RAID (and what kind of RAID) is right for you.

In general, I think it's a good idea not to RAID the system drive (save for a RAID1, but that will give you only security in terms of drive failure!). The second thing is that your backup strategy is -- in any case -- way, way too naïve. Remember that cloning is a very, very bad way to back up your data! Any reasonable backup solution should give you the possibility to (i) restore old data and (ii) go back in time until you have a working system again. If you clone a borked drive, then your backup will be a borked system. Once you've cloned it, there is no way to go back to a working state!

This may not be important for things that don't change much (like backing up the system drive minus data!), but it certainly makes it useless as a sole backup strategy for your (fast changing) data. Also, keep in mind that your backup solution needs to have more capacity than the drives you're using, i. e. if you build a large RAID volume, you'll need an even larger one to back up to. Any storage solution has to include a proper backup solution.

So please post your requirements in details, including the following:
(i) Capacity needed now.
(ii) Growth of data per month or year.
(iii) Types of data with different priorities (e. g. scratch files don't need to be backed up as often or at all; is it video files, pictures, i. e. many small files or fewer larger files?).
(iv) Is there a lot of interdependency between your files or can you separate them into independent projects?
(v) Do you want to be/have to be able to access certain data all the time? If yes, what is their approximate amount (in GB) and growth over time?
(vi) Value of the different types of data.

Just to give you an idea: for larger, fewer files, seek times are less important than through-put. If you can separate your projects well, you can simply charge the customer for a harddrive and store his project on a dedicated harddrive.

[misterdna]

Originally Posted by OreoCookie 

Allen is right-on: without giving us specifics of what you need in terms of capacity, it's impossible to tell you whether a RAID (and what kind of RAID) is right for you.

I hate to say this, but I haven't really been asking you guys to tell me what type of RAID is right for me. I was really trying to learn why some of you said not to boot from a RAID, because a lot of you had said that as a matter of fact, but didn't explain why.

I participated in a few RAID-related threads here (maybe not enough for you guys to remember me?), asked some questions, had some recommendations made, etc. I will probably opt for one of the RAID set-ups that was presented and discussed by some of you. It didn't seem like I needed to present an idea that had already been discussed.

However, the one thing I didn't understand was simply, why everyone rejected the RAID boot idea immediately as if it were common knowledge that it's a really bad idea to boot from a RAID with their new Mac Pro. I never saw anyone say, "it's okay if you have the right backup scenario." No, all I saw was people saying you shouldn't boot from a RAID. Period.

So the details of "why" it is unequivocally a bad idea was what I wanted to know in my most recent posts, particularly since I have found users in other forums (apple, macrumors) who seem blissfully unaware of this edict, and happily are running their new Mac Pros with a single RAID 5 array (perhaps with high risk of impending doom?). Since the ideas for setting up a RAID in a new Mac Pro has been discussed in depth in this and other threads, I was trying to discuss the one thing I didn't see discussed (only stated) -- what are the reasons not to boot from RAID? Thus, my configurations of booting from a single-volume RAID 5 was put out there for the sake of general discussion, just to understand what the big problem with booting from a RAID setup like this is. I really was looking to learn in general terms, not get help from you in determining if this set-up was right for my situation.

Make sense?

Before I order my hardware, I will submit my intended set-up ideas (an details on how I use my system) to see what you guys think. Likely it will be a copy of one of the RAID configurations I've seen discussed already on this forum.

BTW, when I say "clone" I mean I use SuperDuper! to copy my drive to an external (actually two, that I rotate, so I can keep one off site). I do this so I will always have a bootable copy of my entire drive, so if anything happens to my drive (or my computer), I will always have complete data, apps, etc. and can be back up and working very quickly. Cloning my drive (if I'm using the right word) has saved my ass a couple of times, and I've never been told this is a bad idea. Obviously, it should not be the only backup method. I happen to also do incremental backups to REV drives and use Retrospect (running on an older computer) to copy my active jobs folder to another computer twice a day (each to a separate folder), and copy my entire user folder to another computer once a day. I also burn finished projects to DVD, and have kept most projects completed in the past couple of years on another drive. I feel I am probably overkilling it in the back-up department. But as far as my RAID 5 single-volume scenario being "cloned" (using SuperDuper!) sceneario, it was simply to address that, even if a RAID-boot mystery disaster occurred, it seemed like I would easily be able to recover. And I was hoping someone would address that particular scenario for my general learning, not to address it as the right solution to my particular needs.

Eek, I gotta limit myself to one less-wordy post a day, I'm getting behind on my work.

-R

[ninahagen]

So far I have been able to glean only five negatives to RAIDing the boot drive:

1) firmware update problems (referred to by mduell, but not fully explained)
2) inability to boot if any drive in the array goes down (an obvious fact)
3) a more pronounced need for external backup (but any pro should have it in any case)
4) the increased probabilty of array failure due to multiple HDs and a RAID controller (not much of an issue if backup is solid)
5) more complicated recovery protocol (yes, but I would never try to recover the system in order to reboot. With good backup data, you identify the failed HD, pull or replace it, sweep the rest, reinstall the system from the startup/utility disc, reconfigure your working drives and pull the data in from your backup.) Aside from reconfiguring your array, those are the same steps you would take working off a single HD for system, apps, scratch & data... a little more complicated sure, but if you are even attempting a RAID, you have to be ready for that level.

There may be more I don't know of, but those do not seem that daunting.

If that's the sum of the issues, RAIDing the boot drive seems like a good way to gain extra internal HD capacity and the simplicity of a single volume, since you don't have to dedicate a bay and an HD to system & apps.

[misterdna]

Originally Posted by ninahagen 

So far I have been able to glean only five negatives to RAIDing the boot drive:

1) firmware update problems (referred to by mduell, but not fully explained)
2) inability to boot if any drive in the array goes down (an obvious fact)
3) a more pronounced need for external backup (but any pro should have it in any case)
4) the increased probabilty of array failure due to multiple HDs and a RAID controller (not much of an issue if backup is solid)
5) more complicated recovery protocol (yes, but I would never try to recover the system in order to reboot. With good backup data, you identify the failed HD, pull or replace it, sweep the rest, reinstall the system from the startup/utility disc, reconfigure your working drives and pull the data in from your backup.) Aside from reconfiguring your array, those are the same steps you would take working off a single HD for system, apps, scratch & data... a little more complicated sure, but if you are even attempting a RAID, you have to be ready for that level.

There may be more I don't know of, but those do not seem that daunting.

If that's the sum of the issues, RAIDing the boot drive seems like a good way to gain extra internal HD capacity and the simplicity of a single volume, since you don't have to dedicate a bay and an HD to system & apps.

Thank you, thank you, thank you! I guess I haven't been taking crazy pills. This is directly answering the question I had asked. And your conclusions seem generally on par with my own, which gives me some faith in my own powers of research and reasoning... Do any of you think Nina is missing something?

Thanks again!

[ninahagen]

Just thought of one more potential negative and one more question:

— sierradragon told us that the speed advantages of RAID are optimized at 50% or less of the nominal capacity. Is that unique to RAID arrays? Don't standard single drives have speed differences between optimal and nomimal data loads. I have always just used 60%~70% of nominal capacity as a rule of thumb. If say a normal HD were OK up 2/3 of nominal capacity, and a RAID were only OK up to half, your boot volume/stripe would have to be proportionally bigger.

— question: if misterdna head straight off a cliff and do boot from a RAID volume, is it better to make a formal partition of the volume for Leopard, and if so, how big to keep it running snappy?

[mduell]

Originally Posted by ninahagen 

So far I have been able to glean only five negatives to RAIDing the boot drive:

To that I'd add: More chance of problems after an abrupt power outage (without UPS, after UPS is exhausted, when the powersupply goes up in smoke, etc) due to write caching and the lack of atomicity.

Originally Posted by ninahagen 

Just thought of one more potential negative and one more question:

— sierradragon told us that the speed advantages of RAID are optimized at 50% or less of the nominal capacity. Is that unique to RAID arrays? Don't standard single drives have speed differences between optimal and nomimal data loads. I have always just used 60%~70% of nominal capacity as a rule of thumb. If say a normal HD were OK up 2/3 of nominal capacity, and a RAID were only OK up to half, your boot volume/stripe would have to be proportionally bigger.

— question: if misterdna head straight off a cliff and do boot from a RAID volume, is it better to make a formal partition of the volume for Leopard, and if so, how big to keep it running snappy?

The performance curve of a RAID array depends strongly on the RAID level and the controller. It can be flatter or steeper than a typical single drive.

If you're only going to have one array, why bother with more than one volume?

[misterdna]

Originally Posted by mduell 

To that I'd add: More chance of problems after an abrupt power outage (without UPS, after UPS is exhausted, when the powersupply goes up in smoke, etc) due to write caching and the lack of atomicity.

Interesting point, particularly with a software RAID. The wikipedia RAID page says that both these power outage issues are addressed in a hardware RAID: "Most hardware implementations provide a read/write cache which, depending on the I/O workload, will improve performance. In most systems write cache may be non-volatile (e.g. battery-protected), so pending writes are not lost on a power failure." The Mac Pro Raid Card has a 72 hour battery.

And in regards to atomicity, wikipedia mentions: "While the battery-backed write cache may partially solve the problem, it is applicable only to a power failure scenario."

Obviously, the RAID card back-up is not a guarantee that the RAID won't be corrupted due to power failure. But taking a step back, aren't the issues you mention related to all RAIDs, not just when booting from a RAID? Or do you see a synergy that occurs when RAIDing the boot drive that intensifies the risks you mention?

Bottom line is, I'm still not seeing why a RAIDed boot drive going down is of much more consequence than a RAIDed data drive going down. Either way, you reset the RAID (replacing drives if necessary) then clone from your backup. So what puzzles me is, I haven't heard people saying "don't RAID a data drive," but I've certainly heard "don't RAID a boot drive." I still don't think there's been information provided that makes one seem so very much worse than the other. But I'm still all ears...

I genuinely appreciate the indulgence in this subject -- it was worth whining about in a few post to finally get the discussion going. I will have to join the other forums just to reference this discussion to others who had the same question I did.

[mduell]

The battery on the cache helps, but it's no silver bullet. The life of the battery degrades over time, power outages last longer than the battery runtime, people don't necessarily return to their computers immediately to turn them on, etc.

The difference between your boot volume and your data volume is that you can work without your data volume.

[misterdna]

Originally Posted by mduell 

The difference between your boot volume and your data volume is that you can work without your data volume.

If my data is gone, I'm screwed. The computer not booting isn't the end of the world for me, I have other computers I can always work from. So as long as I have a functional backup, the difference between losing the boot & data vs. just the data doesn't seem like a huge deal in my case.

Anyway, I do thank you again for helping me understand the negatives of this option. While the headaches and risks of booting from a RAID don't seem as terrible as I expected, I understand why people tend to prefer booting from a dedicated drive. And now that I feel I understand the negatives, I can try to assess the positives and then decide if a single volume RAID 5 configuration is worth considering.

All comments (even those that I said were off the topic I was pursuing) have been instructional -- it's amazing to have access to so much knowledge and advice.

-Rob

[SierraDragon]

Originally Posted by misterdna 

So as long as I have a functional backup, the difference between losing the boot & data vs. just the data doesn't seem like a huge deal in my case.

Except that OS and data on the same array can facilitate a scenario where all is lost including the backup.

Originally Posted by misterdna 

I hate to say this, but I haven't really been asking you guys to tell me what type of RAID is right for me. I was really trying to learn why some of you said not to boot from a RAID.

Sorry, but you are not paying attention. You say you have done your homework (e.g. Wiki on RAID) but the statements above say otherwise. All RAID arrays are not the same and are totally situation-dependent! The issues you raise are not simply about RAID per se.

Thus, my configurations of booting from a single-volume RAID 5 was put out there for the sake of general discussion, just to understand what the big problem with booting from a RAID setup like this is. I really was looking to learn in general terms, not get help from you in determining if this set-up was right for my situation. Make sense?

NO it does not make sense because "your situation" has not been specified. How many times must I say write it down, the devil is in the details, etc.? "RAID" means redundant array of independent drives. Some arrays in some scenarios are appropriate. Other arrays in different scenarios are inappropriate. Like I have repeatedly said, it is all about the specifics. Without specifics "now that I feel I understand the negatives" is not happening even if you think it is. Ongoing generic discussion is ludicrous, you can get that from Wiki; until you lay out specifics (previously described) I for one am done commenting.

-Allen Wicks

[misterdna]

Originally Posted by SierraDragon 

Sorry, but you are not paying attention. You say you have done your homework (e.g. Wiki on RAID) but the statement above says otherwise. All RAID arrays are not the same and are totally situation-dependent! The issues you raise are not simply about RAID per se.

NO it does not make sense. How many times must I say write it down, the devil is in the details, etc.? "RAID" means redundant array of independent drives. Some arrays in some scenarios are appropriate. Other arrays in different scenarios are inappropriate. Like I have repeatedly said, it is all about the specifics. Ongoing generic discussion is ludicrous; until you lay out specifics (previously described) I for one am done commenting.

-Allen Wicks

Sorry Allen, but while I typically get so much from your posts, the past few days the tone of the discussion between you and me has felt argumentative, and I'm just not here to argue with anyone.

I have wanted to understand a basic tenant that has been thrown around the Mac Pro RAID discussions, and I feel I have now achieved the better understanding I was searching for. I am not quite sure why my approach appears to rub you the wrong way, but quite frankly, I feel slightly attacked by your insistence that I'm not paying attention, and by your calling my approach to learning "ludicrous."

I'm simply trying to coming to my own understanding of all the posts I'm reading before getting to the details of configuring my system... And if I'm going to be attacked, probably best for me to steer clear of posting much else on this board, and just make my own judgments based largely on posts on this board (many of them being your posts).

[SierraDragon]

Originally Posted by misterdna 

...I have wanted to understand a basic tenant that has been thrown around the Mac Pro RAID discussions...

No disrespect is intended. The point is that lacking a specific scenario there is no "basic tenant." Insisting on understanding a basic tenant that without specifics does not exist is fruitless (and I suppose that I become frustrated for my failure to communicate). E.g. in some scenarios I would consider the OS on an array to be an excellent solution. But it is all situation-dependent.

-Allen Wicks

[misterdna]

Originally Posted by SierraDragon 

No disrespect is intended. The point is that lacking a specific scenario there is no "basic tenant." Insisting on understanding a basic tenant that without specifics does not exist is fruitless. E.g. in some scenarios I would consider the OS on an array to be an excellent solution. But it is all situation-dependent.

-Allen Wicks

By reading all the posts discussing internal RAID configurations for the new Mac Pro (here and in other forums), there were a large number of posts by highly-ranked commenters saying "don't boot from a RAID." This was always said as a blanket statement (a "basic tenant" if you will), not as if it were a case-by-case recommendation. I wanted to understand the details of this basic tenant, to understand why a whole subset of potential Mac Pro RAID set-ups seemed unworthy of consideration.

As for wikipedia's entry on RAID, I could probably read it every day for a month and get something new out of it on every read. Same for many of these threads. It may not be as easy for me to absorb the more technical details as it is for people with more technical experience than I have, so perhaps information that's obvious to you was a bit too technical for me to grasp. The bottom line is, after reading all these sources, I didn't understand anything that explained the encountered blanket dismissals of RAID boot volumes in the new Mac Pro. Hence my attempt to have a focused discussion specifically on that general issue.

I've spent literally dozens and dozens of hours over the past few months reading and rereading information about RAIDs (and other Mac Pro configuration information) over the past few months. Last week I began using a spreadsheets to compare different RAM, drive, RAID and Mac Pro speed possibilities and associated prices, to help compare the prices of various configurations (using owc as the source for prices on drives and RAM). Perhaps I don't have as much of a clue as I think I do, but I'm certainly trying.

[ninahagen]

Originally Posted by mduell 

To that I'd add: More chance of problems after an abrupt power outage (without UPS, after UPS is exhausted, when the powersupply goes up in smoke, etc) due to write caching and the lack of atomicity.

The performance curve of a RAID array depends strongly on the RAID level and the controller. It can be flatter or steeper than a typical single drive. If you're only going to have one array, why bother with more than one volume?

Thanks so much Mark, for both of these points.

On the first, wouldn't using hardware RAID and a good power strip with surge protection mitigate the worries around atomicity?
Is there any way to completely eliminate this issue?

On the second, I am presuming hardware RAID0 (with the Apple card) is going to be flatter/better than the other types of RAID. Is that right? Does Sierra Dragon's estimate of 50% optimal data load, still hold for a RAID0?

[ninahagen]

Originally Posted by misterdna 

I'm just not here to argue with anyone.

I understand misterdna's itch to understand this thing, since I suffer from it too. I also understand Allen's frustration at explaining a thing over and over and not getting through.

I have a suggestion. Could you, would you, misterdna, accept Allen's invitation, and begin working out your system and the RAID issues together? I am not saying Allen is necessarily right or wrong about it being situation specific. It seems there would be a way to get at it theoretically, but since he is one of the best and most experienced posters on this forum, and he prefers to frame the issues in context, why not use the chance to get his help? You will have to ask these questions eventually anyway, so you are only accepting timing that isn't your ideal... concurrent vs. sequential. I am certain you will find him (and oreo and mark) very helpful.

I will start by giving my situation, and you can follow it up with yours.

I just used Oreo's requirements, interspersing my answers below...

(i) Capacity needed now.

150GB active data. I also have a number huge archival files (nearly a TB) that virtual never change, but those can be kept on DVDs.

(ii) Growth of data per month or year.

For stuff I want to keep ready at hand, say another 150GB a year.

(iii) Types of data with different priorities (e. g. scratch files don't need to be backed up as often or at all; is it video files, pictures, i. e. many small files or fewer larger files?).

Top priority are large picture files (hi res tif scans and camera RAW data) often a full 1GB each... sometimes in batches of 200 images. So, though I might have 200GB of files I am working on at once, after they are edited, the huge masters are put on DVD, and only 200 x 1 MB jpgs remain in the HD. Photoshop and Illustrator with some Indesign work. Not using aperture now, but considering it. A fair amount of batching.

(iv) Is there a lot of interdependency between your files or can you separate them into independent projects?

Can be seperated. Love Oreo's idea of dedicated HDs for clients.

(v) Do you want to be/have to be able to access certain data all the time? If yes, what is their approximate amount (in GB) and growth over time?

already answered above.

(vi) Value of the different types of data.

Many many thousands of $.

**********

My proposed solution is:

4 x 450GB 15k Cheetahs in RAID0 (with Apple RAID card)
Limit the total data to optimal level...
System, apps, scratch & data all on one volume
8GB RAM
Standard vid card

1TB Time Capsule as backup (stowed in a waterproof container in a fireproof safe every night).

***********

2 possible alternates:

— use the extra optical drive bay for a system+apps drive. Does anyone know if I could fit two HDs in the optical bay?

— and possibly whittling down the RAID array to 3 drives, and having a dedicated scratch as well.

************

Oreo brought up a major issue re: backup... the ability to go backward in time.

With a 1TB time capsule, how much room should I leave for weekly versioning. Say the RAID0 turns out to have a flatter curve and I can use 60 of nominal drive capacity.

With the 3 x 450 RAID option, I would have 1350GB nominal, 810GB to play with, say 800GB to round it off. That leaves 200GB for versioning. Something tells me that is not enough.

*************

OK, mister dna, your up!
Alternate

[OreoCookie]

misterdna, you do your research upside down. Don't start looking at configurations, investigate your requirements first. It's pointless to talk about configurations, unless you know your requirements. perhaps you don't even need a RAID for storage (e. g. by using separate harddrives for clients). Wikipedia or any other good source on RAIDs won't be able to help you unless you know what you need.

[ninahagen]

Hi there Oreo,

Would you be kind enough you help a little with my RAID project? Not only will it halp me lots, but give mr. dna a positive example of the synergistic approach. I have been so occupied here, I haven't had time to work on this since a few months... but the need for an update is getting more urgent.

Many thanks as always...

Nina

[misterdna]

Originally Posted by OreoCookie 

misterdna, you do your research upside down. Don't start looking at configurations, investigate your requirements first. It's pointless to talk about configurations, unless you know your requirements. perhaps you don't even need a RAID for storage (e. g. by using separate harddrives for clients). Wikipedia or any other good source on RAIDs won't be able to help you unless you know what you need.

People! You are all acting like I haven't already posted a number of questions and comments in RAID threads over the past few months. I've interacted with all of you regarding RAIDs well before this particular discussion about booting from a RAID.

You'll find I started using this forum to research RAIDs back in November, originally by asking Allen what his reasons were for his plans to install a RAID in his system (in a closed thread, Mac Pro Waiting Blues).
http://forums.macnn.com/65/mac-pro-a...7/#post3537288

Then, on January 22nd, after a lot more reading, I posted basics about what I do on my Mac (BTW, "avocation" means hobby -- I think I threw Nina off with this term):
http://forums.macnn.com/65/mac-pro-a...0/#post3584849

Besides participating actively in these threads, I've also sat back and followed every thread that discusses RAIDs, paying particular attention to those who use their Macs as I do.

But PLEASE do not respond yet, I've got a lot more details to give you, but I also have a very busy schedule today. I'll try to updated this post later with details about my data, back-ups, ideas for my new Mac Pro system (which are really just ideas Allen and mduell have tossed around), etc.

[SierraDragon]

Originally Posted by misterdna 

By reading all the posts discussing internal RAID configurations for the new Mac Pro (here and in other forums), there were a large number of posts by highly-ranked commenters saying "don't boot from a RAID." This was always said as a blanket statement (a "basic tenant" if you will), not as if it were a case-by-case recommendation. I wanted to understand the details of this basic tenant...

It is not a basic tenant, even if your reading implies that it is. As long as you insist that it is conceptually a basic tenant, understanding cannot occur.

(Note that I will shortly suggest to Nina that IMO her proposed RAID with OS is probably not a good idea for the specific setup she proposes. That does not make it a basic tenant that RAID/OS is a bad idea in every possible scenario. E.g. if MPs had 10 bays I might RAID0 two fast drives for the OS and apps and weekly clone to a third drive for OS/apps backup; then I might have 3-4 drives RAID0 for data+scratch and 2-3 drives to back those up. The specific scenario with ten drives available does not exist for me but it does for some. Time Machine and its complexity would further complicate, but I just wanted to make one basic possible example scenario.)

As for wikipedia's entry on RAID, I could probably read it every day for a month and get something new out of it on every read...

Me too! That is why I routinely reread that white paper and recommend that others do the same.

-Allen Wicks

[misterdna]

Originally Posted by ninahagen 

OK, mister dna, your up!

To continue from above.

First, I will remind you all of the initial information I provided (remember, "avocation" is intended to mean "hobby" not "vocation"):
http://forums.macnn.com/65/mac-pro-a...0/#post3584849

**********

Here's my info that parallels Nina's post:

(i) Capacity needed now.

195GB active data (this is the size of my user folder, it includes active jobs and 4GB of finished projects that need to be archived)
260GB total on drive (includes OS, apps & data... eek, and the trash has 8GB in there -- I will empty it after I clone my drive tonight!)


(ii) Growth of data per month or year.

For stuff I want to keep ready at hand, I'll estimate 50GB a year max.


(iii) Types of data with different priorities (e. g. scratch files don't need to be backed up as often or at all; is it video files, pictures, i. e. many small files or fewer larger files?).

Top priority is whatever project I'm working on. Total working size of individual projects (mostly print projects consisting of mainly Photoshop and Illustrator files) commonly range from 25MB to 20GB. The biggest projects are up to 50GB, but that's atypical. I can often slim down the archived data of these projects by about 30% by deleting outdated files in the project's folder (I do a lot of incremental saves that can be tossed once a project goes to print).

Another top priority is personal photos (76GB and growing) and music I've recorded (guessing 30GB and growing). Growth rate is included above, see: (ii).

And then theres video... I don't create a ton of video projects (hobby, not work), but they can range from 5GB-50GB easily, though much of that data can be scrapped if I'm truly finished with a project, since I still keep the source HDV or DV tapes. And since this is still just a hobby, it's not so vital.


(iv) Is there a lot of interdependency between your files or can you separate them into independent projects?

Mostly independent projects, but it's not rare to have to access pieces from time to time.


(v) Do you want to be/have to be able to access certain data all the time? If yes, what is their approximate amount (in GB) and growth over time?

About 150GB at all times, growth rate listed above (ii) is estimated at 50GB/year max.


(vi) Value of the different types of data.

Photo collection, recorded live music, and other personal projects have enormous sentimental value.
The cost of a print job "disappearing" from my computer the day it's due to go to print is hard to estimate, but I'll say the bigger print projects top out with a theoretical non-performance lawsuit of $25K-$150... Though I'm probably being paranoid, and the numbers are hopefully much lower. Heck, I could get hit by a car the day a project goes to print and still be sued for non-performance. That's why I pay a lot for E&O insurance.

**********

And here are more details:

1) I currently use retrospect from a remote computer to:
-create an incremental backup to REV drives (I committed to REV drives when they were new, hoping prices would come down for media. They haven't, so I may change to another media type). this is so I can (theoretically) access any file that's ever been on my main drive... ever...
-duplicate my active jobs folder twice a day, to two different folders on my remote computer, so if my main drive dies, I won't have lost too many hours of work.
-duplicate my user folder once a day to a different drive on the remote computer

2) I use two 500GB externals to clone my drive to, rotating one or the other offsite once in awhile (I didn't mention this on the original post).

3) I burn finished work projects to DVD (or CDs in early years), and have done so for 10 years. The discs have remained remarkably stable.

4) I have transferred all my finished work projects for the past couple of years to my 500GB scratch drive (which, from Allen's posts, I've learned keeping so much data on there is hurting the performance of the scratch drive). I do this for redundancy, and it also makes it easier than searching for the right DVD archive if I need to access the occasional old file. I also keep an OS on this scratch drive so I can boot from it when doing various maintenance on my main drive.

5) When working, I often have several applications running at the same time, such as: Photoshop, Illustrator, Mail and FIrefox (if not more, like Palm Desktop and Bridge (CS3)).

6) I plan to start using Aperture

7) I also use audio software, 3D software, Garage Band, may buy the light version of Logic, and a bunch of lesser known apps.

8) For my video hobby (at least it's just a hobby for the time being) I use most of the FCP suite, including Motion, Soundtrack, DVD Studio Pro and Compressor. I have a consumer HD camera (Canon HV20) that requires transcoding to get 24p from the direct captures.

9) I also expect to move into Blu-ray at some point (I have no clue when), possibly as a way to back-up data and also to burn HD Blu-ray movie discs.

**********

My proposed solution is: Based on posts and other reading, my current plan is to order a Mac Pro with base drive and RAM (I'm not sure on speed, right now I'll guess 3 GHz), 8800 GT, RAID card, Applecare. Then I'll likely order (from OWC) 6x2GB RAM modules to add to the 2x1GB modules, 4x1TB drives and a Raptor. I don't need anywhere near that much room, I'm looking more for speed. The plan is to boot from the Raptor, and then to use some RAID configuration with the 1TB drives, keeping them relatively empty to get the most speed from them.

Deciding on the RAID configuration for those 1TB drives seems to be the toughest decision. Mduell has mentioned RAID 1+0, but it's not supported by the Apple RAID card. So that leaves me with choosing some configuration of RAID 0, 5, or 0+1 (JBOD doesn't fit my needs). I also have to decide if I'll have a dedicate scratch drive (perhaps a scratch RAID 0 two drives), or if instead I should RAID all 4 1TB drives (as mentioned before to RAID 0, 5, or 0+1) and skip the dedicated scratch drive. I am slightly leaning towards: boot from Raptor and make the 4 1TB drives a RAID 5.

As for backup, I would buy two 1TB externals that I would rotate, just as I do currently with my two 500GB externals. I don't see needing to backup a full 1TB in the near future. If I were to go Raptor boot then I would make two partitions on the external, one for the boot drive, and one for the data volume (be it RAID or not). I will also continue to run Retrospect remotely on to provide several redundant back-up of my data. I would probably use one of my old 500GB externals as an additional repository of completely jobs, and also keep it bootable, so I can start from this drive for disk maintenance, etc.

Note that I wouldn't mind having two optical drives for burning DVDs (movie and data) at twice the rate, and for copying discs, etc.). Thus I have a very small feeling of not wanting to put the Raptor in the 2nd optical drive slot.

**********

I'm open to corrections and suggestions, but would hope the tone can be friendly, not condescending or critical. Yes, I'm getting a little sensitive to what sometimes feels like harsh responses to what I thought were reasonable posts.

[misterdna]

Originally Posted by SierraDragon 

It is not a basic tenant, even if your reading implies that it is. As long as you insist that it is conceptually a basic tenant, understanding cannot occur.

I think you put your finger on our failure to communicate. I've never meant to insist that it is a valid tenant! I've been saying that people have implied it is a basic tenant within their posts, and I've wanted to understand why the implication was made.

Originally Posted by SierraDragon 

IMO putting the OS on a RAID array is simply not a good idea.

-Allen Wicks

When someone makes a recommendation that I don't understand, and the recommendation is relevant to a decision I am trying to make, I ask "why?" so I can have this relevant information at my disposal when I make my final decision. I simply was looking for some understanding of the relevant principles involved in the recommendation, and I figured people who made the statements would happily provide the reasoning behind their opinions (just as people are happy to explain "why" they like a particular drive or application). I never would have asked the question of "why" if I were able to figure it out on my own (which I tried to do without success through research before posing the question on this board). And I never would have fathomed that this question would rile anyone up and be labeled as "absurd."

And at this point I truly ask that we exit this particular discussion, as it feels like it's devolved into a completely semantic argument that isn't benefiting anyone. Perhaps I'm too sensitive, but it stresses me out to see my seemingly innocuous posts trashed over and over, and to constantly feel the need to defend where I'm coming from. I'm wanting to learn from you guys, not argue with you or be put down by you.

[mduell]

Why is everyone being such a dick in this thread?

Originally Posted by misterdna 

If my data is gone, I'm screwed. The computer not booting isn't the end of the world for me, I have other computers I can always work from. So as long as I have a functional backup, the difference between losing the boot & data vs. just the data doesn't seem like a huge deal in my case.

By "gone" I mean "needing to be recovered from backups." Not all of us have the luxury of multiple work computers set up and ready to go. It's like RAIC!

Originally Posted by ninahagen 

On the first, wouldn't using hardware RAID and a good power strip with surge protection mitigate the worries around atomicity?
Is there any way to completely eliminate this issue?

On the second, I am presuming hardware RAID0 (with the Apple card) is going to be flatter/better than the other types of RAID. Is that right? Does Sierra Dragon's estimate of 50% optimal data load, still hold for a RAID0?

Neither hardware RAID nor a power strip does much of anything about the atomicity issues. I'm not really familiar with high end disk systems, but I assert you can buy your way to atomicity; beware, it probably makes a new house look cheap.

I haven't seen any profiling of Apple's RAID card; given the price, Apple's usual markup, and Apple's performance claims, I'd be comfortable with recommending going to 75% full with any RAID level.

Originally Posted by ninahagen 

I will start by giving my situation, and you can follow it up with yours.

150GB active data. I also have a number huge archival files (nearly a TB) that virtual never change, but those can be kept on DVDs.

For stuff I want to keep ready at hand, say another 150GB a year.

Top priority are large picture files (hi res tif scans and camera RAW data) often a full 1GB each... sometimes in batches of 200 images. So, though I might have 200GB of files I am working on at once, after they are edited, the huge masters are put on DVD, and only 200 x 1 MB jpgs remain in the HD. Photoshop and Illustrator with some Indesign work. Not using aperture now, but considering it. A fair amount of batching.

Can be seperated. Love Oreo's idea of dedicated HDs for clients.

Many many thousands of $.

**********
My proposed solution is:

4 x 450GB 15k Cheetahs in RAID0 (with Apple RAID card)
Limit the total data to optimal level...
System, apps, scratch & data all on one volume
8GB RAM
Standard vid card

1TB Time Capsule as backup (stowed in a waterproof container in a fireproof safe every night).

***********

2 possible alternates:

— use the extra optical drive bay for a system+apps drive. Does anyone know if I could fit two HDs in the optical bay?

— and possibly whittling down the RAID array to 3 drives, and having a dedicated scratch as well.

************

Oreo brought up a major issue re: backup... the ability to go backward in time.

With a 1TB time capsule, how much room should I leave for weekly versioning. Say the RAID0 turns out to have a flatter curve and I can use 60 of nominal drive capacity.

With the 3 x 450 RAID option, I would have 1350GB nominal, 810GB to play with, say 800GB to round it off. That leaves 200GB for versioning. Something tells me that is not enough.

*************

150GB now / 20% rule of thumb = 750GB
600GB after 3 years / 75% rule of thumb = 800GB

So a two drive stripe of those 450s is fine for data (with 100GB for estimation error). That leaves us with two bays, one for OS/apps/housekeeping (iCal, Mail, etc) and one for scratch. Each of those drives could be 300GB or even 150GB.

You could put a 1TB PATA (not sure about using those extra 2 SATA ports when you're using SAS in the bays) in the optical bay for Time Machine. Add another 1TB FW800 or eSATA external to clone the TM drive to. I really don't see a reason to suffer the performance or cost of Time Capsule when you're going to keep it in a safe most of the time. This also makes it easy to upgrade your TM target: physically upgrade one side (internal or external), clone, physically upgrade other side, clone.

Remember, Time Machine is not a versioning system; versioning systems don't delete your oldest versions unless you tell them to! Versioning should be done outside of Time Machine with CVS, Subversion, or one of the media management packages.

Archiving on to DVDs is a good idea; you may also want to archive on to external hard drives (perhaps one per major customer and one for all the rest) in addition for faster recovery (assuming the hard drive archive still works... that's what the DVDs are for). Apply the usual rules of thumb for each client's drive.

[Bwa]

This whole "dont boot from RAID" thing is a lot of BS as far as I'm concerned. I've been doing it for years and I used to work at tech companies developing enterprise storage appliances--thousands of customers running tens or hundreds of our devices--that boot from RAID. My day job is another enterprise storage start-up, but software instead of hardware this time.

I'd be more concerned about making sure that if your RAID is degraded, you get an alert of some kind. Apple's software RAID doesn't do that. Apple's software RAID didn't used to support RAID volumes other than RAID-1 for the boot volume due to the issue of the kernel image being spread out. Maybe that is fixed these days with their software RAID.

I have 4 internal drives and 4 external drives on my Mac Pro, all in various RAID configurations.

The main concern to be aware of is that if you trust your data to a hardware RAID, and the controller fails, you're going to need that same controller to get your data back online without doing a recovery from a back-up. For this and many other reasons, do not buy into RAID as a back-up system--it's a fault tolerance system to keep your system online, and in many configurations, it is a performance system.

[OreoCookie]

@nina
You should definitely separate storage drives from scratch drives and system drives. It will slow down operation when an app wants to access some resource files while simultaneously accessing the image files.

So I think it would make more sense to do the following:
(i) A dedicated system drive. You could get a Raptor for instance, although if you really want to throw money at it, you can get a 15k SAS drive. I'm not sure whether you will see as much of a speedup as when investing in the correct data drives.
(ii) A 2 or 3-drive RAID0 scratch array. You don't need Raptors, for instance, there are 7.2k drives with larger sustained data throughput (which is what you need). 15k SAS drives give you extra performance, but at a price (RAID controller + SAS bonus).
(iii) A storage drive, e. g. one of the latest 1 TB drives gives you decent performance and storage space. Remember that you will work primarily with your scratch RAID and not with your storage drive.
(iv) Backups: now if I were you, this is where I'd spend money on. Forget about the RAID card and the SAS drives, invest here first.
If you need more than 5 TB of storage capacity, you need a professional RAID solution. A Time Capsule won't do it (think of it as something for consumers). If you definitely want something small and portable, a Drobo is a better alternative, although it's still a consumer product.

I think you should make your backups onto a professional RAID system, but these things aren't cheap. Now, there are some options, depending on what drive interface you prefer and how much storage space you need. You have solutions like the WiebeTech RT5e on the low end (with PATA interface for the harddrives -- which doesn't mean it's slower, but that you probably will have limited upgrade options in the future), but you can get an XRaid-like RAID box (the XRaid since has been discontinued, Apple has an officially supported replacement, but I forgot the name, though) that attaches via FiberChannel to your Mac. WiebeTech offers a professional 8-drive RAID with SCSI interface between your Mac and the RAID. The cheap WiebeTech will set you back about 2.5-3.5k, the XRaid-like box will cost you about 10k including sufficiently many drives. If you have several workstations, you can use an older (fast) Mac as a server and save all of your groups work on it. A dual/quad G5 is more than plenty for that. If you use gigabit ethernet, you can get decent throughput for nightly backups.
(v) In addition to that, bill your clients for an external drive for each of your projects. Drives are cheap and they don't need to be fast. Drives are faster and if your clients wants all his files, give him `his' harddrive and be done with it. One of our members (I think art_director) is doing that. This way, you have all your data in at least two places.

The advantages:
(i) Your computer will just work even if your RAID0 crashes! No (important) data is lost (the data storage drive is independent of the RAID0!).
(ii) You can access system drive, scratch drive and storage drive without the drives interfering with each other. This gives you extra performance.
(iii) Spend money in a smart way. If you need more storage, you can change the storage drive in your Mac without touching your RAID0 or your system drive. Better invest in a good backup solution, don't rely on consumer hardware for the most important point of failure!
(iv) You can clone your system drive only (which does not really change), but do real backups with your data. (Clones are not backups.)
(v) You put the point of failure where it belongs: to the data you don't necessarily need the most. You can have a cheap (but equally fast) software RAID0 without much concern for data safety.

[SierraDragon]

Originally Posted by OreoCookie 

@nina
You should definitely separate storage drives from scratch drives and system drives...
...(ii) A 2 or 3-drive RAID0 scratch array.

Nina only needs 100-200 GB for scratch. Dedicating 2-3 drives for 200 GB is very wasteful. And I am not convinced that having scratch and images/work files on the same RAID0 array is much of a slowdown.

My intent for my MP (not enough drives to RAID yet) is OS/apps Drive A; RAID0 scratch/images Drives B/C/D; on site Backup Drive E. All internal. The RAID will probably have 2 partitions, the smaller of which will be 200 GB for scratch so scratch can be routinely erased (unless someone convinces me that partitioning the RAID0 is a really counterproductive idea).

-Allen Wicks

[OreoCookie]

@misterdna
I think you don't really need a RAID. If you work with video, modern drives usually offer sufficient bandwidth to satisfy your needs (especially since this is more of a hobby).

I think you would benefit more form getting a dedicated scratch drive and -- if you really want -- a dedicated system drive.

Two solutions
Two-drive solution
(i) Get a 1 TB drive for storage and the OS.
(ii) Get a scratch drive, e. g. a 150 GB Raptor might already fit the bill. Keep in mind that you get a Raptor only because of its seek time which helps when you have many small files that you want to access. There are drives that have a larger data throughput than the Raptor.

Three drive solutions
(i) Get a Raptor as system drive.
(ii) Get a 1 TB drive for storage.
(iii) Get a scratch drive, e. g. a Raptor or a fast 7.2k drive.

So let's see:

You have 260 GB of active data + OS and your data growth is 50 GB/year. Let's triple that, just to be super-safe. Assume you won't upgrade for three, four years. That is 260 GB + 4x150 GB = 860 GB. All of the data you will generate in the foreseeable future will fit on a 1 TB drive. I don't see why you have to buy 4 GB or raw storage (i. e. 3 GB in a RAID5, 2 GB in a RAID10), it's a waste.

A scratch drives accelerates matters as it is independent of your other `data storage' drive. Copying folders back and forth should be much faster -- faster than copying files from a RAID0 onto itself! Even if you make a scratch RAID0, but if you tend to work with smaller files, you won't benefit as much from the increased throughput.

A RAID1 protects you against drive failure (hardware!) -- which is not necessary if you do backups regularly, so I think this is not really necessary.

Keep in mind that if you need additional storage space, you can always add at least 1 more drive, at most 3. If it turns out that in two years, you really do need more storage, you can either get a 1.x TB harddrive which is available then or make a RAID of several 1 TB drives.

Backup
Now let's turn to your backup solution. I think you do need to get a 1 TB drive for your data, especially if you back up your video files, too, and would like to keep several versions of it. Your backup solutions sounds pretty good, you use Retrospect (I suppose you do incremental or differential backups, so you have several versions of your files available), but even a Time Machine drive in addition to that will help a great deal! Being able to recover old versions of files has helped me a great deal. Just make sure to exclude your video scratch files!

If you want to replace your REV drives, consider a tape drive. They're not cheap (especially those with useful capacities), but worth it. Much better than burning DVDs, unless you're really religious about burning your projects to DVDs!

[OreoCookie]

Originally Posted by SierraDragon 

Nina only needs 100-200 GB for scratch. Dedicating 2-3 drives for 200 GB is very wasteful.

I've added a 3-drive RAID0 for reasons of data throughput. Plus, she could get three smaller Raptors or two larger ones for a similar price, but the number of drives was just motivated by throughput. Nina is working on files that weigh in several hundred MB a piece (or so I remember from earlier discussions), so this is what will help her.

Originally Posted by SierraDragon 

And I am not convinced that having scratch and images/work files on the same RAID0 array is much of a slowdown.

Depends on how you use your system. If you don't use them simultaneously a lot, then yes, you won't benefit. But it can accelerate your work a lot, e. g. if you keep only finished files in storage and copy a lot from your scratch volume to your storage volume and vice versa. I think the main reason they're around today is that there are no 1 TB Raptors for sale

Originally Posted by SierraDragon 

My intent for my MP (not enough drives to RAID yet) is OS/apps Drive A; RAID0 scratch/images Drives B/C/D; on site Backup Drive E. All internal. The RAID will probably have 2 partitions, the smaller of which will be 200 GB for scratch so scratch can be routinely erased (unless someone convinces me that partitioning the RAID0 is a really counterproductive idea).

I wouldn't want my data to sit on a RAID0, unless I were certain I have them somewhere else. Having a RAID0 for storage when you have a scratch volume means to me that you don't need as much speed for your storage volume. If you data fits onto one drive and you can separate your data well (e. g. into projects), you don't need a large RAID0 that holds them all. So even if push comes to shove and your 1 TB is 70 % full, you just pop in another one (or a larger drive which has since become available).

Partitioning a RAID0 does not make much sense to me. Depending on what you do, it can slow things down. Copies from your scratch volume to your storage volume will be slower, no matter the data size (it's like copying files on the same harddrive, the throughput is dominated by the seek time). If you say, you don't do that a lot, you won't notice that drawback. But then again, you won't be copying a lot from your storage volume to your scratch volume, so that speed (for the storage volume) is not as important. Adding that to what I've said above, I think you'll be better off getting a single non-RAIDed drive for storage and a scratch drive.

[ninahagen]

Originally Posted by SierraDragon 

Nina only needs 100-200 GB for scratch.

Is that really all I need? Even with a batch of 100 x 1GB files? If so, I am really pleased, because I would like to save the space.

Originally Posted by SierraDragon 

Dedicating 2-3 drives for 200 GB is very wasteful. And I am not convinced that having scratch and images/work files on the same RAID0 array is much of a slowdown.

That is a key question. Would the extra speed from having the scratch volume across 4 drives (instead of 2) outweigh any potential interaction-based slowdown?

Originally Posted by SierraDragon 

My intent for my MP (not enough drives to RAID yet) is OS/apps Drive A; RAID0 scratch/images Drives B/C/D; on site Backup Drive E. All internal.

Will you be putting os/apps drive or the backup drive in the optical bay? It seems that heat would be a factor in that decision. Are you going to use SAS? mduell told me about a great wire harness you can use to connect SAS drives via PCIe cards.

Any idea if 2 drives would fit in the optical bay?

Originally Posted by SierraDragon 

The RAID will probably have 2 partitions, the smaller of which will be 200 GB for scratch so scratch can be routinely erased (unless someone convinces me that partitioning the RAID0 is a really counterproductive idea).

I really hope a member can speak to partitioning. I don't know enough to make a decision. I had only thought of that in order to minimize the kind of interactions/slowdowns oreo alluded to.

[SierraDragon]

Originally Posted by misterdna 

I think you put your finger on our failure to communicate. I've never meant to insist that it is a valid tenant! I've been saying that people have implied it is a basic tenant...

All we can go by is what you write. Our failure to communicate may have something to do with the fact that your precise previous words were:

I have wanted to understand a basic tenant...
I wanted to understand the details of this basic tenant...

My statement "IMO putting the OS on a RAID array is simply not a good idea" was in response to benz, the original poster, who said

yes, by using the RAID utility in macpro Mac OS, u can easily create volumes, and i create a boot up volume ( sized 100G) containing Mac OS, and other space for data volume , both volumes are in the RAID set at RAID 5. but u need a RAID card installed in the mac pro, which costs ~800USD, so it's hardware RAID, but one point i think it lacks is "e mail notification of the failed drive"

My response to benz's specific scenario was that IMO putting the OS on a RAID array was simply not a good idea.

So in any event, hopefully it is now clear that I opine that sometimes the OS on RAID makes sense and other times it does not, totally depending upon each individual scenario.

For a scenario with 4 or 5 drives total IMO the OS and apps should be on a single non-RAID drive. Such a drive can be backed up by cloning weekly or even less often.

RAID0 maximizes speed and capacity and significantly increases risk of failure. Backup of RAID0 should occur daily or even more often.

Backup of hundreds of GB of data from a large array is not an instant process, nor is it always seamless and perfect. A few corrupted bytes in a TB of image files is generally no big deal, because original image files always exist elsewhere off site anyway. A few corrupted bytes in the OS can be a HUGE deal, because when the OS is corrupt the entire multiple-TB RAID can be corrupt.

-Allen Wicks

[SierraDragon]

Originally Posted by ninahagen 

Is that really all I need? Even with a batch of 100 x 1GB files? If so, I am really pleased, because I would like to save the space.

You should have available something like 30-60 times the file size, and scratch only applies to the files you have open and active. So 200 GB is actually overkill. But PS assigns multiple scratch disks anyway, so if for some reason your Primary scratch needs did exceed 200 GB the overflow would simply go to the 2nd and 3rd drives assigned for scratch.

Will you be putting os/apps drive or the backup drive in the optical bay? It seems that heat would be a factor in that decision. Are you going to use SAS?

No, SAS is too expensive. When I get to 5 drives the backup will be in the optical bay.

Any idea if 2 drives would fit in the optical bay?

I doubt it.

-Allen Wicks

[ninahagen]

Originally Posted by SierraDragon 

A few corrupted bytes in a TB of image files is generally no big deal, because original image files always exist elsewhere off site anyway. A few corrupted bytes in the OS can be a HUGE deal, because when the OS is corrupt the entire multiple-TB RAID can be corrupt.

-Allen Wicks

THAT is gold. This reason alone convinces me to separate the OS/apps from the data. Particularly after what oreo said about cloning a borked drive, with the consequence that the (ahem) "backup" is then borked, and may be useless for retrieving data in a meltdown. Just imagine going to all that trouble to set up a blazing RAID volume, back it up with an expensive elegant machine, and find on crashing that you had lost all your data. I can give up my data bay up to eliminate that doomsday scenario. Unless you have great versioning within your backup system (expensive and huge), and can smoothly go ever backward until you can find the pre-borked version, keeping the OS/apps drive separate from the data drive now makes perfect sense.

This clear reason however does not contraindicate putting the system & apps in their own dedicated RAID volume, presuming one were to accept the other (to me, lesser) risks inherent to RAID.

Several questions arrive from this new awareness (thanks to all here for helping me come to it):

— What are the potential downsides of putting the scratch volume together with the data volume?
— What are the potential downsides of putting the scratch volume on the OS/apps drive/volume?

I am thinking now of:

System Volume = Bays 1 & 2 (SAS drives)
Storage Volume = Bays 3 & 4 (450GB Cheetah SAS drives in RAID0, self-limited to 700GB of data)
Scratch Volume = 2 drives in Optical bay or elsewhere in the machine (any suggestions here)

—OR—

System Volume = Optical Bay (SAS drives)
Storage & Scratch Volumes = Bays 1~4 (4 x 450 = 1.8 TB nominal, self-limited to 1TB data and 300 scratch)

—OR—

System & Scratch Volume = Bays 1 & 2 (SAS drives)
Storage Volume = Bays 3 & 4 (450GB Cheetah SAS drives, self-limited to 700GB of data)

Which would you choose?

[OreoCookie]

No, it doesn't, but both, speed- and capacity-wise, I don't see a reason to use a RAID0 or RAID5 for your system volume.

[ninahagen]

Originally Posted by OreoCookie 

No, it doesn't, but both, speed- and capacity-wise, I don't see a reason to use a RAID0 or RAID5 for your system volume.

Is the OS undemanding enough on a modern drive, that it will function just as well (in terms of real world usage) with a 7200rpm single drive as with a multidrive RAID0 array? That is not a rhetorical question... that seems it really could be the case.

Which of the three configurations do you guess would be the fastest for heavy PS work? (they can be tweeked to reflect the single system drive option, except in the case of integrating scratch and system into one volume, in which case RAID0 would make a lot of sense... even if the system doesn't need the speed, the scratch would surely benefit.

Thanks,

Nina

[misterdna]

Thanks for the detailed response. I do have some questions/comments based on your comments:

Originally Posted by OreoCookie 

@misterdna
I think you don't really need a RAID. If you work with video, modern drives usually offer sufficient bandwidth to satisfy your needs (especially since this is more of a hobby).

Hmmm, interesting. I initially didn't plan on getting a RAID, but other people on this forum said (in a general sense) the Mac Pros were so fast that the drives were now the bottleneck. But maybe people only notice the bottleneck for video stuff, not Photoshop work?

I think you would benefit more form getting a dedicated scratch drive and -- if you really want -- a dedicated system drive.

Honestly, drives are so cheap, I won't even blink an eye before getting plenty of drives so I can have dedicated drives for each function (boot, data & scratch). So you think the separate drives will actually give me better performance (particularly with large Photoshop files) than a Raptor boot drive and a (yes, wasteful) single-volume 4x1TB RAID 5 but no dedicated scratch drive? I guess I'm trying to discern how much your recomendations are centered around saving me money vs. maximum performance.

(ii) Get a scratch drive, e. g. a 150 GB Raptor might already fit the bill. Keep in mind that you get a Raptor only because of its seek time which helps when you have many small files that you want to access. There are drives that have a larger data throughput than the Raptor.

This is an area I'm shaky about. I don't know how much quick seek times will speed me along compared to quick data throughput. Also, many people talk about a RAID 0 array with two fast drives for scratch. You think that's extreme overkill?

I don't see why you have to buy 4 GB or raw storage (i. e. 3 GB in a RAID5, 2 GB in a RAID10), it's a waste.

Assuming you mean 4 TB, etc., right? Anyway, since I was assuming I would go with a RAID card, and because the drives are relatively inexpensive, the potential "wasted" extra space didn't bother me. I was also under the belief that some 1 TB drives perform really well, particularly with low amounts of data on them (Allen and others talk about the excellent performance levels of drives that don't have much data on them). So I really was looking at the 4 big drives for RAID speed, with the subtle benefit of having plenty of headroom. That was my thinking...

Copying folders back and forth should be much faster -- faster than copying files from a RAID0 onto itself!
Even if you make a scratch RAID0, but if you tend to work with smaller files, you won't benefit as much from the increased throughput.

Well, here's the thing. I'm really looking for my new system to kick ass when I'm working on huge Photoshop files or very complex Illustrator files, or moving back and forth between these apps. And saving the huge Photoshop files can take awhile. These are the times I get a little bogged down on my G5 (Dual 2.7). So I would really like to have a system that screams in these situation. In your opinion, what set-up will help me in these situations?

Backup
Now let's turn to your backup solution. I think you do need to get a 1 TB drive for your data, especially if you back up your video files, too, and would like to keep several versions of it. Your backup solutions sounds pretty good, you use Retrospect (I suppose you do incremental or differential backups, so you have several versions of your files available), but even a Time Machine drive in addition to that will help a great deal! Being able to recover old versions of files has helped me a great deal. Just make sure to exclude your video scratch files!

If you want to replace your REV drives, consider a tape drive. They're not cheap (especially those with useful capacities), but worth it. Much better than burning DVDs, unless you're really religious about burning your projects to DVDs!

I actually lived by DAT tape systems (DDS3, DDS4) for a number of years, and they were always SCSI. But Macs moved to USB and Firewire, yet these DAT tape systems I had come to live by didn't seem to change from SCSI. So I though I was changing with changing times, stopped using tapes and switched to REV drives (which I've felt like a sucker for doing). Do you know what tape-based systems are Mac & Retrospect friendly and not overkill? If not, I can look around.

As you may be sensing, I'd rather error on the overkill side with my new Mac Pro rather than feel I spent the money on the new computer and didn't spend a few more dollars to really make it fly. I mean, I have my limitation (like all 4 SAS drives is too expensive), but I'm not worrying about spending a few hundred more bucks to get bigger drives than I really need, or an extra internal drive. Since I realize budgetary issue influence recommendations, I figured I would mention my perspective on this. If I were really watching my dollars, I'd just keep using my G5, which still works fine. I figure that, if I'm going to buy a new system, I want it to kick-ass when I'm working on and saving 1 GB Photoshop files. And if the extra bucks help slightly speed my work AND give my video hobby a generous speed gain, then it's worth it to me.

Thanks again for your input!

[SierraDragon]

Originally Posted by ninahagen 

— What are the potential downsides of putting the scratch volume together with the data volume?

IMO probably not all that problematic. Obviously having each component (sys, apps, scratch, data) on its own independent striped RAID array would be theoretically ideal but for most of us that gets ridiculous in drive costs and backup complexity.

— What are the potential downsides of putting the scratch volume on the OS/apps drive/volume?

Absolutely unacceptable. The way scratch works it is a significant performance hit for scratch to be on the same physical drive as apps/OS.

Which would you choose?

Among your 3 options I choose:
System Volume = Optical Bay (SAS drives)
Storage & Scratch Volumes = Bays 1~4 (4 x 450 = 1.8 TB nominal, self-limited to 1TB data and 300 scratch)

I would RAID0 the storage/scratch volumes.

Backup, however, was not specified in your options, and chosen backup protocol can change everything.

-Allen Wicks

[SierraDragon]

Originally Posted by misterdna 

-CS3 software, and deal with multiple-page projects with working files that at times can be in excess of 20 GB.
-newest versions of FCP and Motion
-(i) Capacity needed now.
195GB active data (this is the size of my user folder, it includes active jobs and 4GB of finished projects that need to be archived)
260GB total on drive
-(ii) Growth of data per month or year. For stuff I want to keep ready at hand, I'll estimate 50GB a year max.
-Top priority is whatever project I'm working on. Total working size of individual projects (mostly print projects consisting of mainly Photoshop and Illustrator files) commonly range from 25MB to 20GB. The biggest projects are up to 50GB
-Another top priority is personal photos (76GB and growing)and music I've recorded (guessing 30GB and growing). Growth rate is included above
-video... I don't create a ton of video projects (hobby, not work), but they can range from 5GB-50GB easily
-I plan to start using Aperture
-Note that I wouldn't mind having two optical drives for burning DVDs (movie and data) at twice the rate, and for copying discs, etc.). Thus I have a very small feeling of not wanting to put the Raptor in the 2nd optical drive slot.

My proposed solution is: Based on posts and other reading, my current plan is to order a Mac Pro with base drive and RAM (I'm not sure on speed, right now I'll guess 3 GHz), 8800 GT, RAID card, Applecare. Then I'll likely order (from OWC) 6x2GB RAM modules to add to the 2x1GB modules, 4x1TB drives and a Raptor. I don't need anywhere near that much room, I'm looking more for speed. The plan is to boot from the Raptor, and then to use some RAID configuration with the 1TB drives, keeping them relatively empty to get the most speed from them.

I took the liberty of summarizing some of your needs commentary above so I could see it all in one place. There may be some inconsistencies but that does not matter because IMO the gist is clear enough. Bold/underlined emphasis is mine.

Your needs are very similar to mine, except that I need more capacity than you do and you seem to have a looser budget. Also you would like two optical drives while I will just replace the Superdrive with a Blu-ray when feasible and cost effective. Also I suspect that you are better organized than I am (an issue that applies to backup protocols as well as to how fast hard drive storage needs grow).

I am repurposing a 2006 MP: replaced RAM with 4x2GB=8GB matched RAM from OWC in the top riser. Added an HD2600XT card because the 8800GT is not yet available for the 2006 MPs.

The intent (so far I only have 3 of the planned drives in place) for my MP is OS/apps Drive A; RAID0 scratch/images Drives B/C/D; on site Backup Drive E. All internal. The RAID will probably have 2 partitions, the smaller of which will be 200 GB for scratch so scratch can be routinely erased (unless someone convinces me that partitioning the RAID0 is a really counterproductive idea).

Drive A apps/OS is cloned weekly or less often to an off-site external. RAID0 B/C/D is backed up daily to Drive E internal. Drive E is backed up to an off-site drive weekly. For me "off-site" is simply in the back of my Yukon, readily accessible, but unlikely to share any catastrophes like fire/theft with my office.

I believe that you are likely underestimating storage growth needs. Added on your other uses, DSLR photography and the joys of Aperture will very likely cause more rapid growth than you forecast.

I disagree with Oreo when he says RAID is inappropriate for your scenario. Aperture alone will show benefit from RAID0 in your setup. And, drive speed is very dependent on how full the drive or array is. A RAID0 as it gives you speed also gives the speed of always having the array be less full than a single drive would be.

My suggestion (which includes Aperture's demanding hardware needs that exceed CS3's needs) to you is:

• Base 2008 MP but if your budget tolerates moving up in clock speed go for it.
• Replace the Apple RAM with 4x2GB=8GB matched RAM from OWC in the top riser and move Apple's 2GB RAM to the bottom riser. Personally I consider the benefits of matched RAM worthy, so IMO we should add in sets of 4 matched DIMMs. Since 4GB-sized DIMMs are pricey now I will wait until prices fall a bit and then add 4x4GB RAM.
• 8800 GT card but note that the stock HD2600XT is a pretty good card already.
• Drive A sys/apps, stock Apple or replace it with a Raptor for substantial speed benefit.
• Drives B/C 1 TB each, RAID0 for scratch and apps.
• Drive D 1 TB for on-site daily or more frequent backup of B/C.

Backup protocol similar to mine above but feel free to modify however you see appropriate for you. Note that I am ignorant as to the needs of video; I suggest the FCP forum for that.

-Allen Wicks

[misterdna]

Ha, no arguing on this one. Thanks for taking the time to go through this for me. Only three questions/comments immediately come to mind:

Originally Posted by SierraDragon 

The RAID will probably have 2 partitions, the smaller of which will be 200 GB for scratch so scratch can be routinely erased (unless someone convinces me that partitioning the RAID0 is a really counterproductive idea).

I realize that "scratch" means a few things to me. With CS3 applications, you designate a scratch drive, and it seems all scratch files are invisible, and I assume the invisible files "go away" when the file or application closes (I don't "know" any of this stuff, this is just what I think happens). Then, from my somewhat limited FCP suite experience, it seem "scratch" means visible directories with files you can very much access and manipulate, and need to delete manually at the appropriate time. So I guess I just want to be sure I understand what you mean be erasing the scratch. I guess I am also curious how Aperture writes scratch files (visible, invisible, or?). Feel free to correct me if I'm off base with my beliefs on how the different apps handle "scratch" files.

I disagree with Oreo when he says RAID is inappropriate for your scenario. Aperture alone will show benefit from RAID0 in your setup.

So, seems like you are saying that a RAID as you described, containing scratch and data would give me more overall speed than having physically separate (non-RAID) data and scratch drives. I can see that opening and saving large PSD files would be quicker with a RAID 0 than just a single drive.

• Replace the Apple RAM with 4x2GB=8GB matched RAM from OWC in the top riser and move Apple's 2GB RAM to the bottom riser. Adding 6 GB loses the benefits of matched RAM.

Actually, what I had said was I wanted to add 6x2GB modules (12GB), which would bring the total to 14GB. I interpreted this Barefeats test as saying that having all 8 slots filled was the way to get the fastest RAM performance (my 14GB plan is listed):
Memory Tests: "2008" Mac Pro
What do you think?

Thanks again for taking the time to give me your recommendations!

[SierraDragon]

Originally Posted by OreoCookie 

... wouldn't want my data to sit on a RAID0, unless I were certain I have them somewhere else...

I fully agree. In my case original image files are independently backed up to DVDs before any editing, and the RAID0 is backed up at least daily.

Originally Posted by OreoCookie 

...Having a RAID0 for storage when you have a scratch volume means to me that you don't need as much speed for your storage volume...

?? I do not really understand that statement. But in any event scratch is mainly relevant to Photoshop. Most folks also use other apps (e.g. Aperture) that have no scratch protocol.

Originally Posted by OreoCookie 

...If you data fits onto one drive and you can separate your data well (e. g. into projects), you don't need a large RAID0 that holds them all. So even if push comes to shove and your 1 TB is 70 % full, you just pop in another one (or a larger drive which has since become available)...

Except that RAID0 is significantly faster with no loss of total capacity, and one gains the added benefit of having the array be less full.

Originally Posted by OreoCookie 

...Partitioning a RAID0 does not make much sense to me. Depending on what you do, it can slow things down. Copies from your scratch volume to your storage volume will be slower, no matter the data size (it's like copying files on the same harddrive, the throughput is dominated by the seek time). If you say, you don't do that a lot, you won't notice that drawback. But then again, you won't be copying a lot from your storage volume to your scratch volume, so that speed (for the storage volume) is not as important. Adding that to what I've said above, I think you'll be better off getting a single non-RAIDed drive for storage and a scratch drive.

Yeah, that is the dilemma. I learned long ago that partitioning is most often not beneficial. Since Aperture my Photoshop usage keeps declining, so scratch becomes less and less relevant. And PS is much less hardware-demanding anyway, enough so that my optimization is based solely around what is best for Aperture. However I have a similar consideration with Aperture: Aperture has a Library and separate Master image files. I am not sure of what drives configuration is optimal, but I do know that I want the Library on a RAID0.

Thanks much for your commentary! You have convinced me to forget about partitioning the RAID0.

-Allen Wicks

[mduell]

Originally Posted by OreoCookie 

(ii) A 2 or 3-drive RAID0 scratch array. You don't need Raptors, for instance, there are 7.2k drives with larger sustained data throughput (which is what you need). 15k SAS drives give you extra performance, but at a price (RAID controller + SAS bonus).

I disagree. The usage pattern for scratch is highly random, not sequential reads, and you don't need much capacity here. A single Raptor or 15k SAS (depending which cost/performance arena you're in) is the appropriate disk for scratch; RAID0 only if your scratch files are very large (>25% of the scratch drive).

I think some of us may be using the word scratch differently: to some it means the disk you point Photoshop to, for others (Oreo, apparently) it's where you put your work in progress. I'd use a single high-RPM drive for Photoshop scratch, and a fast array for work in progress/online (not archived).

[OreoCookie]

mduell, Nina is working on large files, that can be a few hundred MB to 1 GB in size. Throughput will be a lot more important than access times. Generally, you're right (and my advice would have been different as well), but in this instance, I think, throughput is more important than random access times.

@SierraDragon
You're right that it depends on how people use their scratch drive. I was actually assuming people would use it to store files that pertain to the project the user is currently working on.

Regarding your comment that a RAID0 is significantly faster for linear reads, yes, you're right. For Aperture, it would make a difference whether your storage volume sits on a single drive or a RAID0, but if you use a scratch volume and -- say -- Photoshop or Final Cut, then it's not as important. It's tricky, different solutions for different applications

misterdna works on a lot of smaller files -- thus access times will become more important. If you use Photoshop, the speedup won't be as significant as in Aperture. If you save a 25 MB file in half a second or quarter of a second won't be as important as if your image project needs 10 seconds to load or just 5. But he doesn't use Aperture yet and Aperture projects tend to explode only if he shoots RAW (in my case, shooting RAW triples the size from about 1.5-3 MB/image to 7-9 MB/image).

On the other hand, the probability of failure of a RAID0 is much higher so that I'm sceptical to advise it to anyone unless I know the data is backed up daily onto reliable media. In your case, I know that you do, but I'm hesitant to advise it as a primary mean to store data to others. Without a storage device in the background that can cope with loss of a drive, I wouldn't recommend it. Cloning your data does not cut it, it doesn't give you the possibility to recover older data (data before a failure, for instance).

Even so, it depends quite a bit on what you're going to use your computer for the most. Running Photoshop is usually a piece of cake for today's machines (unless your name is Nina and you work with huge image files )

[SierraDragon]

Originally Posted by OreoCookie 

...Aperture projects tend to explode only if he shoots RAW...

IMO every photog with a camera capable of modern RAW capture should be using it. In addition to post-process editing, RAW captures much more image data. Personally I shoot "RAW+JPEG." The Nikon D2x RAW files are ~20 MB and the JPEGs 3-5 MB each. The good news is that CF cards are now cheap. Top quality Sandisk Extreme IV 8 GB CF cards are only ~$180. It is not uncommon for a shoot to be 5-10 GB with some much larger, so it is easy to see how quickly mass storage needs can grow.

Better organized folks (perhaps mrdna) may temper growth a bit by offloading projects but I like having full access to all images always, because I build ads that may use pix or portions of pix from long ago. Our discussions here may lead me to using some of my existing external drives as client-specific image repositories, I have not figured that out yet.

-Allen Wicks

[misterdna]

Originally Posted by SierraDragon 

Better organized folks (perhaps mrdna) may temper growth a bit by offloading projects but I like having full access to all images always, because I build ads that may use pix or portions of pix from long ago. Our discussions here may lead me to using some of my existing external drives as client-specific image repositories, I have not figured that out yet.

-Allen Wicks

I'm really only organized with work projects. I'm horribly organized when it comes to my personal stuff. I too like having my entire (personal) photo gallery available to me (in iPhoto), but it could probably be cut into 1/3 or less if I weeded out the crap. However, none of my cameras are near the resolution of yours, so my photo data simply isn't growing as rapidly as yours.

Although I have a lot of back-up going on, it's really just paranoia combined with throwing a lot of methods at combating the risks of losing data that makes you think I might be organized. I clone to have any easy fix if my drive dies (and to keep an offsite copy of everything). I save to DVD because it feels like a good way to archive projects (and it's a 10yr habit). Saving completed jobs to my 2nd drive started as a paranoid move, but now I always do it when I save my project DVDs, and I love having the easy access to so many old jobs. Running retrospect on a 2nd computer with a few hard drives makes it really easy to automatically create a lot of redundant copies of my working data and also give me an incrementally archived back-up... I just gotta make sure all the setting are right, make sure the remote computer is waking up when it needs to and function properly.

BTW, Allen (or anyone), I'm unclear on how to interpret your RAM recommendation vs. the BareFeats test results (below). Keep in mind I don't really know much about RAM issues.

Originally Posted by SierraDragon

• Replace the Apple RAM with 4x2GB=8GB matched RAM from OWC in the top riser and move Apple's 2GB RAM to the bottom riser. Adding 6 GB loses the benefits of matched RAM.

I wanted to add 6x2GB modules (12GB), which would bring the total to 14GB. I interpreted this Barefeats test as saying that having all 8 slots filled was the way to get the fastest RAM performance (my 14GB plan is listed):
Memory Tests: "2008" Mac Pro

[mduell]

Filling the second four slots increases bandwidth but it also increases average latency, which is already pretty poor because they're FB-DIMMs.

[misterdna]

Originally Posted by mduell 

Filling the second four slots increases bandwidth but it also increases average latency, which is already pretty poor because they're FB-DIMMs.

Hahaha. Obviously I haven't done homework on this, because you're talking above my level of understanding. I understand the concepts of bandwidth and latency, but not how they apply to RAM and my computer. I guess I don't get which is the priority, higher bandwidth or lower latency. And I'm not even sure I've ever realized Mac Pro RAM is something called FB-DIMMs nowadays, and definitely had no clue that there were latency issues.

Bottom line, are you saying Allens 10GB recommendation (4x2 + 2x1) indeed trumps my attempt to follow the BearFeats 14GB set-up (6x2 + 2x1)? If so, I'm glad I asked rather than blindly following a chart that someone linked to in another thread.

I don't think I'm up for learning about RAM after spending so much time trying to learn RAIDs!