Harddisk RAID question(and new comp discussion)

[RobbieThe1st]

So, as most of you know, I am building myself a new computer. I have a AMD Phenom 9850 Black Edition CPU and a ASUS Crosshair II Formula motherboard ordered from Newegg.

Now, I need to finalize my new harddisk setup(in the meantime, I shall be using what I currently have(2X250GB raid 1) for testing purposes).
I am looking for one terabyte of usable space, with decent access/read/write times and some redundancy. With these criteria, I see two options:
1. A Raid-5 setup using 3 500GB drives.
2. A Raid-1 setup using 2 1TB drives.
I will be using the raid controller on my motherboard, so that may weigh in on my options.

Now, harddisks to populate these raids:
For the Raid-5 setup, I have been thinking either a WD Caviar SE16 WD5000AAKS or a WD Caviar RE2 WD5000ABYS which both have great reviews.
I *had* been thinking about Seagate 7200.11 drives, however on newegg people are saying that it is kind of a crapshoot as to whether the one you get works or dies. Also, the benchmarks are horrible.

For the Raid-1 setup, I have not done nearly as much research, however the two drives I see as being good possibles are the SAMSUNG Spinpoint F1 HD103UJ and the WD Caviar Black WD1001FALS.

The advantages of the raid 5 setup would be better read/write times, although on the downside it doesn't provide quite as much reliability(33% of the drives can fail versus 50%), and *can* be CPU-intensive. I also wouldn't be able to simply take a single drive out, connect it to a USB adaptor and that to my laptop and then play videos and such off it.

With the raid 1 setup, the read times may be good, write may be worse however, the reliability is good, it is not CPU intensive, and I could take one drive and connect it to a USB adaptor... but for the same quality of drive it may be more money, and the boot times may be slower.


So, now I ask you, the reader:
First, have I concidered everything?
Second, what would be my best option?


-RobbieThe1st

[ShadesFox]

I think that the best option is to get a 1 TB disk and forget the RAID.    If you really insist on RAID, go for the RAID 1.  RAID 5 is horribly slow on write and read, so you are essentially sacrificing performance for cost, losing only 1 disk on an N disk array.  RAID 1 you lose 1 disk on a 2 disk array, but with a smart implementation you can gain on read and lose little on write.  On that note, your MB will probably not have a smart implementation of RAID 1.  Most MB RAID 'controllers' are worthless piles of junk that do little more then inform Windows that there is a 'RAID array' and the real RAID logic is in software.

[Tapewolf]

I'm not 100% sure why you're gunning for RAID here.  If speed is the concern, okay.  If you're using it to try and protect the data, think long and hard about it before committing yourself, and make sure you order a matched pair of RAID controllers.  On-board RAID is a bad idea unless the data is basically valueless and 100% replaceable (e.g. swap, /tmp, /usr, Windows pagefile).  If you're serious, get a matched pair of controller boards and keep one as a spare.

What often seems to happen is that the RAID module uses its own proprietary disk format and when the controller dies or the motherboard has to be replaced, your new one is incompatible and the disks are unreadable.  This has happened to my father and several others beside (and I hate to say I told him so, but I did warn him...) so if I do come across as a doom-monger, there is some justification.

Also remember that RAID won't protect you against rm -rf /, or otherwise corrupting the the data.  Either way, make sure you have a separate backup.

[llearch n'n'daCorna]

Uh...

RAID-5 is faster on read than write; it's faster than RAID-1 for read, but slower on write, since it has to calculate the parity before writing. More disks makes for a faster read, so RAID-5 comes into it's own after 4 or 5 disks; but the overhead for writes remains the same.

RAID-1 is identical to a single drive for write, but faster for read, since it can balance the reads across the spindles.

I don't know where you got your information from, but it appears to be somewhat incorrect.


If you're storing data in a RAID array, you will NOT be able to split the array and use the disk elsewhere. Not unless you're using software raid, and the software is also on the other machine you're connecting it to, and setup will be a bitch, and when you reconnect the drives, it'll moan about how one has changed and the other hasn't - if you're using windows, just mounting the drive will change it, since the access times will all update. You can get around that if you're careful in Linux, but with the defaults, the same problem arises.

And you're still not going to have the ability to just plug it in and make it go.


As Tapewolf says, you probably want to think carefully about why you want to use RAID - the sole purpose of RAID is to allow you to lose hardware and not data; it doesn't replace a decent backup plan, since it doesn't cover user error, but it does mean you can lose a single hard drive. The problem is, if you buy a stack of identical hard drives, chances are they'll end up being from the same batch, shipment, and store... which means they'll all die at roughly the same time.


Personally, I prefer to keep system and data separate - in your situation, I'd be using the dual 250's for the system, and throw in a couple of 500GB or 750GB drives for data; probably RAID-5, but possibly not. I've heard poor reports from fellow admins about the 1TB drives; apparently a lot of failures from some of the manufacturers, but since I wasn't in the market, I didn't pay a lot of attention. I'll look it up for you at some later time...

[Tapewolf]

The problem is, if you buy a stack of identical hard drives, chances are they'll end up being from the same batch, shipment, and store... which means they'll all die at roughly the same time.

That is something I forgot to mention.  There was a very interesting story I read about on slashdot a month or so ago which is a good example of this.
Basically, the firm the guy was working for had an array of 24 units.  One day, one of the disks died.  When they swapped the new one in, the act of rebuilding the array pushed another unit over the edge and it failed too.  If memory serves, they eventually lost 19 of the 24 units that night.  I'll have to see if I can find it again.

Personally, I prefer to keep system and data separate - in your situation, I'd be using the dual 250's for the system, and throw in a couple of 500GB or 750GB drives for data; probably RAID-5, but possibly not. I've heard poor reports from fellow admins about the 1TB drives; apparently a lot of failures from some of the manufacturers, but since I wasn't in the market, I didn't pay a lot of attention. I'll look it up for you at some later time...

Windows seems to love putting data on the system drive, to my supreme annoyance.  Lots of Oblivion savegames and other things are winding up on the volume which I'm not backing up.  If anyone knows a way to force the users to be held on a different volume, I'd be interested to know.

*EDIT*
And I'll bear than in mind about the Terabyte drives.  I've been tempted by them.

[llearch n'n'daCorna]

... let me find the email and confirm that it _was_ the 1Tb drives, and which ones, before you take that advice as gospel. ;-]

[Tapewolf]

... let me find the email and confirm that it _was_ the 1Tb drives, and which ones, before you take that advice as gospel. ;-]

I'm always a bit leery of brand-new storage tech anyway.

Also:  http://hardware.slashdot.org/comments.pl?sid=1003425&cid=25461693

[llearch n'n'daCorna]

Ah, found it.

It's the 1.5TB Seagate drives, see http://techreport.com/discussions.x/15863 ...

[Tapewolf]

It's the 1.5TB Seagate drives, see http://techreport.com/discussions.x/15863 ...

Ahh.  I was wondering if it was the 1.5TB disks, actually since those are the newest and the straight up 1TB ones have had some chance to be proven.  Now you mention it, I do remember reading about that problem.

[RobbieThe1st]

Uh...

RAID-5 is faster on read than write; it's faster than RAID-1 for read, but slower on write, since it has to calculate the parity before writing. More disks makes for a faster read, so RAID-5 comes into it's own after 4 or 5 disks; but the overhead for writes remains the same.

This makes sense - Although technically, with the write, if you had a good controller, wouldn't it be able to calculate the parity fairly quickly, and therefor wouldn't be much slower on the write than a single drive?

RAID-1 is identical to a single drive for write, but faster for read, since it can balance the reads across the spindles.

I don't know where you got your information from, but it appears to be somewhat incorrect.

If you're storing data in a RAID array, you will NOT be able to split the array and use the disk elsewhere. Not unless you're using software raid, and the software is also on the other machine you're connecting it to, and setup will be a bitch, and when you reconnect the drives, it'll moan about how one has changed and the other hasn't - if you're using windows, just mounting the drive will change it, since the access times will all update. You can get around that if you're careful in Linux, but with the defaults, the same problem arises.

And you're still not going to have the ability to just plug it in and make it go.

I am telling you what I personally have done a half-dozen times: I have taken a single volume of my RAID-1 drive out of my current computer(motherboard is a ASUS P4P800-E Deluxe), connected it to my Rosewill USB sata adaptor, and connected that to my IBM T41 laptop via USB. I was then able to access all of my files on the harddisk.
Admittedly, about half the time when I place the drive back into my desktop and start it up, it notices that the drive has been changed and makes me rebuild it, however the other half of the time it doesn't(once it didn't notice even when I had accidentally changed a few files on the second disk... there were some issues when I tried to read the changed files until I rebuilt the array[which I did immediately after that]).

Now, I don't know if my current situation is simply because of cheap hardware or what, but it worked on my old computer.

As Tapewolf says, you probably want to think carefully about why you want to use RAID - the sole purpose of RAID is to allow you to lose hardware and not data; it doesn't replace a decent backup plan, since it doesn't cover user error, but it does mean you can lose a single hard drive.

This is basicly what I am doing - I tend to be good about not deleting my own files, however I am horrible about doing backups(Yea I know I know... but the fact is, I rarely if ever get around to it), and I am simply worried about my harddisk failing. Fortunately, at least with windows, unless you manage to completely erase the data - it is still there and recoverable.

The problem is, if you buy a stack of identical hard drives, chances are they'll end up being from the same batch, shipment, and store... which means they'll all die at roughly the same time.

This is worrying.

I am planning on buying a spare drive no matter which option I go with, so what if I swap one drive out every month or so - thereby creating different usage patterns & different amounts of run time on each and reducing the chance of a mass breakdown.. would that work?

Personally, I prefer to keep system and data separate - in your situation, I'd be using the dual 250's for the system, and throw in a couple of 500GB or 750GB drives for data; probably RAID-5, but possibly not.

I could use my 250s for the system, that would be doable - As it is now I am 'virtually' keeping my data & system separate using partitions.

Ah, found it.
It's the 1.5TB Seagate drives, see http://techreport.com/discussions.x/15863 ...

It appears to be the entire 7200.11 line from the 500GB and the 1TB also(yea, everything can arrive DOA... but they do have an unproportionate number).


-RobbieThe1st

[llearch n'n'daCorna]

"fairly quickly" is not the same as "just write it" - basically, every single byte that goes to the disk needs to stop and be counted on the way through. This involves a couple of context switches in the CPU of the RAID card, at least, plus a whole bunch of memory moves.

Each of these steps takes time, and the speed you're writing to the disk suffers. A good card can minimise that, but, by design, it's still slower than a single disk. And a "good card" is expensive...


If you can pull a single drive from the array, it's not, technically, hardware RAID-1. It's software. Whether you can do that with your new controller or not is dependant on how the techs have built the RAID controller, and what, precisely, it does.


Oh, and your plan of swapping drives in and out? Slightly more risky than simply buying a drive, waiting a couple of weeks, buying the second drive, waiting a month, and buying the third drive, and simply putting one of them in storage...

[ShadesFox]

I've never seen a PC motherboard with a 'RAID' controller that did hardware RAID.  They are just disk controllers with vague delusions of being RAID controllers.  And I stand by my suggestion that the best thing to do for RAID is don't.

[RobbieThe1st]

I've never seen a PC motherboard with a 'RAID' controller that did hardware RAID.  They are just disk controllers with vague delusions of being RAID controllers. 

Hm... this may be good - I shall test my new motherboard before I buy new disks, but this probably means that the new raid will work like my old one.

And I stand by my suggestion that the best thing to do for RAID is don't.

Well, if you have some better way of providing me some redundancy without me having to back things up, or wait for multiple hours while it backs up.. I would love to hear it.


-RobbieThe1st

[Reese Tora]

Well, if you have some better way of providing me some redundancy without me having to back things up, or wait for multiple hours while it backs up.. I would love to hear it.

Well, what, in particular, is it that you are interested in backing up? if you can narrow down what you need backed up, you can cut backup time down by a LOT.
There's little point in backing up program files or system files (there's a stronger case for system files if you make a lot of settings changes) if you're backing up to restore in a crash, and those makesup the bulk of windows installations if you aren't doing a lot of video, audio or image editing or storing, or collecting installers/doing heavy programming.  If you're simply collecting video, audio,  images or installers, you only ever need to back up that stuff once.

An external hard drive that you take offline when not doing backups is the ideal way to backup everything if your'e a regular consumer, IMO, and a lot of external HDD enclosures come with the ability to auto start at a set time, pull a backup, and shut down until next scheduled backup.  They often come with backup software, too.

Another simple backup procedure would be to set a daily copy task using task scheduler and a batch file to run xcopy.  There are a number of options you can use to fine tune what is copied, and you can even copy based on the archive bit (which windows sets when a file is modified, and one switch will unset on files so copied) so only those files which have been modified since last run will be copied.  If you're not worried about having a long term backup, you could place a second drive in your comptuer and just copy what you want to preserve to the second drive.

[RobbieThe1st]

Well, if you have some better way of providing me some redundancy without me having to back things up, or wait for multiple hours while it backs up.. I would love to hear it.

Well, what, in particular, is it that you are interested in backing up? if you can narrow down what you need backed up, you can cut backup time down by a LOT.
There's little point in backing up program files or system files (there's a stronger case for system files if you make a lot of settings changes) if you're backing up to restore in a crash, and those makes up the bulk of windows installations if you aren't doing a lot of video, audio or image editing or storing, or collecting installers/doing heavy programming.  If you're simply collecting video, audio,  images or installers, you only ever need to back up that stuff once.

An external hard drive that you take offline when not doing backups is the ideal way to backup everything if your'e a regular consumer, IMO, and a lot of external HDD enclosures come with the ability to auto start at a set time, pull a backup, and shut down until next scheduled backup.  They often come with backup software, too.

Another simple backup procedure would be to set a daily copy task using task scheduler and a batch file to run xcopy.  There are a number of options you can use to fine tune what is copied, and you can even copy based on the archive bit (which windows sets when a file is modified, and one switch will unset on files so copied) so only those files which have been modified since last run will be copied.  If you're not worried about having a long term backup, you could place a second drive in your computer and just copy what you want to preserve to the second drive.

I have been thinking things over today(great post - it made me think), and am trying to figure things out. Here is what I have come up with:

First off, I haven't done a huge amount of setting changes windows wise, aside from some common ones: Showing file extentions & hidden files; shutting down any services I won't use; a few edits of various registry values when things don't work as they are supposed to, and things like that - Just enough to make it a hassle to reinstall and get things back to working nicely(not to mention getting drivers reinstalled; now *that* is a hassle). Anyway, this means that it would be nice to have some backup/redundancy for these files, although it isn't critical.

Now, program files wise, the major issue is that most games & such save data in the program's folder. I might be able to setup some sort of sub-folder list of files to backup, and just save whatever savegames/settings data I need in the backup without having to backup the entire program files folder.

Files wise... that is where my space goes(out of my 250GB setup, 182GB is my data folder, and I have only 700 megs free currently). A good portion of that, by space, is video - stuff that I would love to simply stick somewhere safe and delete off my main HD(but don't want to lose). In the same catagory, I have a bunch of installers(22GB downloads folder), which can safely get shoved away until I need one of em.
I also have a relatively large collection of MP3s, which are an interesting problem, as I have been going back through them and sorting/re-tagging them(sometimes adding lyrics), which could screw up my backup.
Next going down the list are images. I have one heck of a lot of images from multiple sources - lots from my camera, which I would love to safely archive, but even more from the Internet(I save anything I come across that is the least bit interesting).
I also have a large collection of saved comic archives(15.4GB currently - 76,345 files), which gets updated with perhaps a megabyte of new ones every single day.
At the bottom size wise, I have a rather large number of saved webpages, scripts, and such.


Now, thats all fine and easy to categorize and backup in theory, however I am horrible about sorting things into proper folders, so I have several different folders filled with a jumble of downloads - Images, webpages, zip files and just about anything you can think of. I have recently started trying to save new stuff to multiple folders, which helps... but I still have a large mass to take care of.

I *may* be able to get away with simply doing proper backup & such for the new files, and just sort the old ones as I have time - But, backup wise, I am going to have to set it up when I install on the new HDs.


So, that is my full situation. Have any suggestions on how I should catagorize/backup/install things?


Oh, also:
I was thinking of using my 2 250GB disks for my main boot disk... but the 1TB(or 500GB) disks are at least a third faster data-transfer wise.. so I think it would be a good idea to keep my windows/linux installs on the new HD.


edit:
Hm... for the videos & other files, is there a good open source program any of you know that I can use to back em all up to a set of DVDs, and get any one file out of the set without having to decompress/restore the entire bundle?


-RobbieThe1st

[Reese Tora]

Well, it sounds like it would be difficult to just extract a few things... 
a couple of options:
you can ceate a full backup and then use partial backups to grab changes, this requires a lot of maintainance, or a third party program, because the best way to do it with the tools in windows is with the built in backup manager, which deosn't do scheduled backups.
One thing with this, though, is that you wouldn't be able to just read the one or two files you needed from the backups (atleast, not without another program designed to do that)
You also would need to restore from the backups in sequence, starting with the oldest, which can be a pain if you've got more than a week's worth.

you could use xcopy, as I've mentioned, but that has a problem backing up some protected system files that windows blocks access to, it has the advantage that you're just doing a massive copy command, though, so all the files are just in the new location without compression.

You can use raid, but that has the risk that multple parts of the system might fail at once and permanently lose data

One thing, most recent games actually put the save files in your named documents and settings sub folder, it's only older games that really violate the seperation of program and data, so you can get away with just setting backup on the documents and settings folder (whioch covers your "my documents" and your desktop)

It's way too late, so I hope I was coherent there.

[RobbieThe1st]

Well, it sounds like it would be difficult to just extract a few things... 
a couple of options:
you can ceate a full backup and then use partial backups to grab changes, this requires a lot of maintainance, or a third party program, because the best way to do it with the tools in windows is with the built in backup manager, which deosn't do scheduled backups.
One thing with this, though, is that you wouldn't be able to just read the one or two files you needed from the backups (atleast, not without another program designed to do that)
You also would need to restore from the backups in sequence, starting with the oldest, which can be a pain if you've got more than a week's worth.

you could use xcopy, as I've mentioned, but that has a problem backing up some protected system files that windows blocks access to, it has the advantage that you're just doing a massive copy command, though, so all the files are just in the new location without compression.

Well, what I am looking for is a simple sequential-backup to DVDs tool. Really, WinRAR does everything I am looking for - It can split my data into DVD-sized chunks, it compresses the data, and I can extract one file at a time, and it can also create a report of the RAR/ZIP with filename/path, filesize, modification time, and CRC. What I need is a "frontend" to winrar which will keep a large text-database of all the currently archived files & their respective archive names(and basically, the DVD they are stored on). It will, on a schedule, look at a list of folders, and compare file modification times/CRCs with the currently stored list, and anything different gets written to a new archive.

(Perhaps also have it so I can set certain folders to be archive & delete - permanently archiving the files, then deleting them from my HD)

This shouldn't be hard to do - I could probably script it using PHP(command-line php) & a batch file, however I can't find proper documentation on WinRAR's command-line options.

You can use raid, but that has the risk that multple parts of the system might fail at once and permanently lose data

Although, you have to admit that(especially in addition to a backup scheme) it is less risky than a single drive.

One thing, most recent games actually put the save files in your named documents and settings sub folder, it's only older games that really violate the seperation of program and data, so you can get away with just setting backup on the documents and settings folder (whioch covers your "my documents" and your desktop)

Bold parts are the important ones here - I currently only have 2 games which do this(Halo and SimCity 4). Also, it isn't just games, but other things as well(like my Trillian IRC & chat logs).

It's way too late, so I hope I was coherent there.

Just enough.

Edit:
Back on topic, what type of harddisk to get, no matter what I decide.
I found another option: WD6400AAKS
review
This one looks like it has good overall performance, at low(11.7 cents/GB) cost.


-RobbieThe1st

[llearch n'n'daCorna]

For the commandline options to winrar, try looking up the options to rar itself...

[RobbieThe1st]

For the commandline options to winrar, try looking up the options to rar itself...

HM... I will look.

So, update:
Today I got my motherboard & CPU, took some memory I had gotten off ebay for it a bit ago, and the PSU out of my current computer, hooked everything up, and it worked. I got it to post, recognized one of my two 250GB HDs, and even start booting it! Windows however didn't like the different processor or something, and so bluescreened halfway through booting up(even in safe-mode), but that isn't surprising, considering the difference between old and new computers.

I just now ordered one of the 640GB HDs. When that arrives, I shall install windows XP 32-bit on it, get things working. Meanwhile, I will have a GPU(Unless I see something wrong with it, a Nvidia 9800 GX2) on order, which I will be able to test with the setup I built using the HD. Another few days after that, a second HD(which I will be ordering) will arrive, and at that point I will attempt to make a RAID-1 setup using the two HDs. If I can take one of the raid volumes and use it via a USB adaptor, I shall keep the raid setup, otherwise I shall go with a single-HD setup and back stuff up more frequently.


-RobbieThe1st

[Tapewolf]

Just out of paranoia, did you run memtest on it to be sure?

[Azlan]

Ah, found it.

It's the 1.5TB Seagate drives, see http://techreport.com/discussions.x/15863 ...

I can officially confirm this, as I have been forced to RMA a number of the Seagate drives that I have had to push back our product dates till end of first quarter 2009.  I have also, while I was in London, received a Seagate 2TB drive for evaluation purposes, along with a solid state drive.  I haven't been too impressed with SSDs, but we shall see.

Anyways, I've only found RAID, for non-enterprise data center and SAN applications, to be useful in a RAID-1 setup for read performance improvements as it pertains to gaming... hardcore pwnage gamers.  Most 'RAID Controllers' on the MB are a bugger to set up and well, everyone else has spelled out in far better type than I would have how this works.  I have to agree with llearch, which I tend to hate to do, but he's just so thorough. 

[RobbieThe1st]

Just out of paranoia, did you run memtest on it to be sure?

Haven't... yet. I, um... forgot completely about it.

Today I was digging out a old machine, one of the vintage of my comp that I had fixed(but for one reason or another, it didn't get used for long, but was still grimy)... I took the case completely apart and washed every single bit of it. I also managed to make two really nice finds: The PSU in it was one we had gotten more recently, a 500w Antec Earthwatts, which had a 24-pin motherboard connector! Yay! I also found a 250GB harddisk of the same model as my two(!). This I got permission, formatted, and installed windows XP PRO(Yay! There was a license that was used with this comp, but it isn't in use now, so I get it), and then installed all the drivers I could find.
It so far seems nice and stable, but I shall be running diagnostics tomorrow(and, if it passes... using it.

So, things are great now, and once my new HD arrives, I should be able to copy my system partition(45GB) using Ranish Partition Manager, and copy + resize my data partition using the same.


-RobbieThe1st

[Zorro]

Really unless you are running a server 2 drives in Raid 0.

Drive #1 is OS Only.  Make as many partitions as you like for multi-boot.

Drive #2 is a Backup of Drive #1 PLUS all Data Files.  Minimum of two partitions.

Ideally there would be a drive #3 to back #1 and #2 to that is stored securely off-site and backed up every night, but that is a Enterprise level security paranoid type thing.