Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

4TB DIY RAID enclosure spanks CalDigit HDPro. And costs < $1500.

Omg, It kills me when I see DIY that emphasize only on cheap and speed. Its like DIYing cars that can only do fast straight line with no airbags. There are 3 major factors I can tell that contribute to a potential unreliable system in the article

1. bad choice of RAID controller
2. soso choice of drive
3. RAID 0 config

For any branded or DIY raid system, there is nothing more important than figuring out the original design house/manufacture of the RAID controller. A lot of the names out there are just rebranding the same controller and selling at higher price. A decent controller makes a whole lot of difference in both performance and reliability. A cheap RAID controller may be functional for sometime but once it drop data or the 'entire RAID' and you are seriously fucked. I personally recommend either Areca or LSI as both are on top of the RAID industry. (and they are only ~200 bucks more expansive)

For point 2, the reason why branded system drives are so much more expansive is because they don't use anything unless its enterprise level... Actually i take that back, some company love profit so much they are wiling to use desktop drives to cut cost and sell at a charming price. Video rendering is a critical application that require 'intensive' read and write over long hours which a normal desktop drive will likely fail after an year. Therefore, most branded system only uses enterprise level drives when selling system that comes with standard 2 year warrenty. If you really want to save the $$ then Hitachi Deskstar probably has the best rated reliability.

For my point 3, 500mb/s at RAID0 LMAO :rofl: , those 500mb/s are only an average figure and if you look at the graph closely, you'll see massive downward spikes which implies potential frame drop on each spikes. Even though with RED, you probably don't require full 500mb/s to remain functional but I am telling you that 500mb/s are not as good as advertised. A 'consistent' read/write are desirable for media application and the best way to tell if it works is to use it with your actual equipment or software over 'long' period of time and repetitively to see if the result is reproducable under 'variety' of conditions(OS/memory/CPU..etc). What is more stupid about RAID0 is that you get zero redundancy which kills the main reason of using a RAID controller in the first place. You might as well just buy a SSD with that much money and save whole lot of complexity. The only people who may find this type of setup applicable are gamers who only wanted the speed improvement and data corruption event can be easy fixed by reinstalling the game+OS.

For chassis, there are a lot of different options as long as you are careful with airflow/heating problem and using a stable and sufficient power supply.

Since these RAID are mostly are being used for serious business instead of high school projects, you take the risk when building a DIY system. Being in this industry (I am a tech at Areca), I've seen people scream/cry/yell/sue...do watever when their entire work for their feature simply "dissapeared"... When such event happens, even Santa can't help you if the original manufacture/vendor simply replys with they are very sorry. Here at Areca, only the most experience RAID data recovery techs are allow to touch failed system because they are the only one with the slightlest chance to perform a successful data recovery under so many different variables. I am not trying to bash people building DIY RAID but a lot of testing and engineering design cocepts are essential when combining so many different hardwares together. After all, you use your best judgement and do not just do something after reading just one article.

Happy RAIDing. :cheers2:
 
DIY RAID

DIY RAID

Long time lurker, first time poster here. Just thought I'd give my 2cents worth on the DIY RAID thing. After much pain in the past (mostly on other people's systems!) using "Fake" RAID cards (ie non hardware RAID) I went with an ARECA 1221X ($860 Au with shipping) and a Sans Digital 8 bay tower (http://www.sansdigital.com/towerraid/tr8x.html) (I paid around $500 AU)with 8 1TB Hitachi drives (about $110 each AU when I bought them, even cheaper now). Running RAID 6 I get about 440 MBread/400 MB write per sec with the array 80 percent full.

I get ridiculously higher speeds with the array empty, but does anyone have their RAID empty? I've tried RAID 0 using the internal bays of my Mac Pro, but the overheads and risks just aren't worth it. That way lies pain. Trust me, I've been there. Clients don't care why the data went away, they just care that it did. $2300 AU is not a lot for that level of storage,
if you live in the US, you'd get it a fair bit cheaper cos you don't have to pay the shipping fees I do (total shipping for all the bits was around $350 AU)
and you have more choice of suppliers also.

Let's be honest, given that RAID stands for Redundant Array of Inexpensive Disks and there's no redundancy whatsoever in RAID 0, shouldn't we call it "AID 0". Coincidentally, 0 AID is exactly what you can expect if you lose data on on of these setups. If you're happy to back up to other external drives from your AID 0, that's great, but isn't that exactly what RAID 5 and 6 do all the time as a matter of course?

I don't even fully trust my RAID 6, I back everything up from it at the end of every days work using Sync Pro to another less expensive RAID 5 box via ESata. Yes, I could get higher numbers if I used RAID 5, but I prefer the comfort of knowing my files are slightly safer.

On the subject of speeds, I used the Blackmagic disk utility, I can't help but notice noone quoting numbers is mentioning where those numbers are coming from, which makes comparisons pretty useless really. Xbench for example gives me much higher numbers for the same array, but I don't think they are all that accurate. Because I'm a nerd, when I first built this array, I spent a few days testing.

One of the things I did was to play uncompressed 10 bit HD out
to my monitor (HP dreamcolor via Decklink with an AJA convertor so that I can actually use the Dreamcolor's engine) and physically ripped a drive out whilst doing so. Not a hitch. Ripped a second drive out and it kept right on playing. Whilst I wouldn't advise doing this as a daily exercise, it does give me an awful lot of confidence in my setup, that I doubt I would have in any of the non hardware RAID solutions.

Given the price of these things has plummeted in the last 2 years, I can't understand why anyone who depends on fast reliable storage would settle for anything less. My setup is a very inexpensive solution for what it is, I looked at all the options available and there was not a lot of difference in price between hardware RAID and port multipliers. Even RAID 0 port multiplier solutions are not that much cheaper, in fact way more expensive when you factor in the inevitable downtime when a drive does die and you need to recreate all the data you lost. Really, we're posting on a forum where we're talking about dealing with footage from a pretty expensive camera (I know, I know, its cheap compared to the other big guns, but hey, I was using HDV and DVCPro before this!) and we're quibbling over a few hundred bucks, maybe a grand. Personally, Once I've acquired footage, I don't really feel the burning urge to go shoot it again......... Your mileage may vary.
 
Last edited:
It's late here, so too sleepy to read the whole thread - are these fast DIY Raid solutions useable with top spec previous generation 17" MacBook Pro with xpressCard/34 slot? Or overkill for a laptop?

Thanks for any thoughts!
 
Thanks for the reply Mark.

Actually what I'd like is an ultimate desktop MacPro and big, very fast editing storage.

Unfortunately I can't afford both at the moment.

I'm happy to struggle with my MacBook Pro doing personal stuff, but want some big fast storage now which, while currently overkill, I can still use when I get enough savings to buy the asskicking MacPro. But it needs to be able to move data at the top speed the previous generation MacBook Pro can connect via xpressCard/34 slot and an expansion card.
 
Well. I've been reading this thread with much interest. Indeed, similar discussions on the same exact topic have been unfolding on several other forums for the past couple of years.... and yet.... no clear unequivocal turnkey solution has emerged.
Just so you know, my time is worth a lot to me so I'm about to put out a challenge for someone to spec out the "Ultimate RAID" for uncompressed storage and will attach a prize to the challenge. It seems to be the only way that a bullet proof spec will emerge that can possibly compete with the much ballyhooed CalDigit boxes.
 
Andrae Palmer said:
I called Enhance Technologies about the UltraStor JS and they actually recommend the LSISAS3801E as the controller for their JBOD. The card is significantly cheaper and faster than the ATTO. I think this is due to the ATTO R380 not supporting Wide Port. The UltraStor JS is Wide Port and uses a LSI SAS Expander Board.
Click to expand...

Hi Andrae- ATTO are high quality though and "proven". Whats the word on these for durability. Reliability? How NEW are they?

Also what is wide port? any links I can read.

Mark P. Great find, thanks!

Everyone else using RAID 0 ... thats just insane with R3D's man on paid jobs and even for personal projects, time is money. Have you guys ever lost footage from a failed drive yet? You will kick yourself for using RAID 0. Raid 5 ( I use 6) is plenty fast for most HD work... and the peace of mind is priceless.

cheers
 
M.Paul El-Darwish said:
...Just so you know, my time is worth a lot to me so I'm about to put out a challenge for someone to spec out the "Ultimate RAID" for uncompressed storage and will attach a prize to the challenge. It seems to be the only way that a bullet proof spec will emerge that can possibly compete with the much ballyhooed CalDigit boxes.
Click to expand...

When you step back and look at the amount of data that we are working with what you find is that there is no magic bullet as a storage solution because we all have different business dynamics.

I'm currently mulling over making a 12 disk FreeNAS system because I need common storage for several computers. This is different than someone wanting 8 high speed disks local to their workstation to do direct renders (I'd say Highpoint card(s), generic cases, and WD Green Power drives for a budget & energy friendly solution).

But previous posters are right to point out that RAID 0 is a loosing proposition. Loosing one disk takes out the whole volume, which leaves you wading through backup tapes to get your too large volume restored (you are baking up to LTO4, aren't you?).

Also, there is the concept of 'Fast Enough', as in 'When I render my CPUs hit 95%, so my disks are... Fast Enough!". A lot of guys throw around GB/s numbers like the disk is the obvious bottleneck in our systems... I have a 3.0 GHz Dual Quad Mac Pro with the Apple RAID Card and 4 1.5 TB Seagate drives in RAID5, and when I start two threads of Redline my CPUs shoot up to +90%, and the disks aren't really phased at all. My system is CPU bound, which is all you can ask of a computer.

My advice:
  • Next time you are running a render, open a terminal window and run 'Top' If your CPU time is somewhere near 90% then disk speed (at your current level if disk usage) is not an issue.
  • Don't store things on disk. Get a HP Storageworks LTO4 tape drive and an ATTO SAS card and offload to tape when you finish projects.
  • Stay away from RAID0; get a dedicated RAID card as the heart of any direct attached solution and do RAID5 or 6.
  • Seagate and WD have quasi-pro drives that are meant for RAID usage. They are a little more costly and probably worth it.
 
Simon Blackledge said:
Andre you'll be looking at either a SAS with poss an expander or one that uses pci-e extender.

Fastest.. well.. there all the fastest :) lol.. to be honest depends on the sas expander and the HBA and the drives etc..

JMR are good. The sonnet is the jmr rebranded.

iStar USA do a 16bay expander
http://www.istarusa.com/storage/sas/dage316u40sasexp.aspx

netstor
http://www.netstor.com.tw/_03/03_02.php?NzQ=

Also Enhance tech do one
http://www.enhance-tech.com/products/ultrastor/RS16_JS.html

and a newish (rebrand) ci design do one now.
http://www.istoragepro.com/ir16saej.html

These require a raid card in the macpro

If you get one with raid built in ( more $$$) you just need a sas hba without raid.

Personally I'd got for a AttoR380 or maybe Areca but prefer atto.

Drives, Hitachi Ultrastars if 2TB. If 1TB Hitachi do a 24x7 which is deskstar but is apparently enterprise/raid capable.


If you go for a pcie version basically you buy a raid card (hba) and put that in the chassis and connect to the macpro with a card that comes with the chassis. so you need a raid card with internal not external connectors.

Though tbh a 8 bay with no expander would probably be fine for Smack*
http://www.enhance-tech.com/products/desktop/E8_MS.html
We run these with an ATTOR380 and at 70% full currently were still getting over 500MBs



hope that helps. anything else DM me or hit me on Wave :)
Click to expand...
 
Last edited:
riggieri said:
I have a PROAVIO enclosure with 8 Hitachi 1TB "generation 2" drives connected to a Highpoint 4322 in a Raid 5 Config. Empty i was seeing close to 700mb/sec write and 750mb/sec write. if you get the Segate 7200.12 hard drives you can probably get over 1GB/sec in Raid 0. Cost me less than $2200. the PROAVIO is a great enclosure. Can play two streams of 2K DPX frames fine. Even when the raid is 75% full. If you put two of these cards in a system, you can even do 4K DPX playback.

Matt Riggieri
Click to expand...


The Proavio is made by Enhance.
Try using LSI-3801e and running RAID Zero or doing a SDA in your O/S.
900 is common. Try some tunning tricks and watch it fly.
 
James Liu said:
Omg, It kills me when I see DIY that emphasize only on cheap and speed. Its like DIYing cars that can only do fast straight line with no airbags. There are 3 major factors I can tell that contribute to a potential unreliable system in the article

1. bad choice of RAID controller
2. soso choice of drive
3. RAID 0 config

For any branded or DIY raid system, there is nothing more important than figuring out the original design house/manufacture of the RAID controller. A lot of the names out there are just rebranding the same controller and selling at higher price. A decent controller makes a whole lot of difference in both performance and reliability. A cheap RAID controller may be functional for sometime but once it drop data or the 'entire RAID' and you are seriously fucked. I personally recommend either Areca or LSI as both are on top of the RAID industry. (and they are only ~200 bucks more expansive)

For point 2, the reason why branded system drives are so much more expansive is because they don't use anything unless its enterprise level... Actually i take that back, some company love profit so much they are wiling to use desktop drives to cut cost and sell at a charming price. Video rendering is a critical application that require 'intensive' read and write over long hours which a normal desktop drive will likely fail after an year. Therefore, most branded system only uses enterprise level drives when selling system that comes with standard 2 year warrenty. If you really want to save the $$ then Hitachi Deskstar probably has the best rated reliability.

For my point 3, 500mb/s at RAID0 LMAO :rofl: , those 500mb/s are only an average figure and if you look at the graph closely, you'll see massive downward spikes which implies potential frame drop on each spikes. Even though with RED, you probably don't require full 500mb/s to remain functional but I am telling you that 500mb/s are not as good as advertised. A 'consistent' read/write are desirable for media application and the best way to tell if it works is to use it with your actual equipment or software over 'long' period of time and repetitively to see if the result is reproducable under 'variety' of conditions(OS/memory/CPU..etc). What is more stupid about RAID0 is that you get zero redundancy which kills the main reason of using a RAID controller in the first place. You might as well just buy a SSD with that much money and save whole lot of complexity. The only people who may find this type of setup applicable are gamers who only wanted the speed improvement and data corruption event can be easy fixed by reinstalling the game+OS.

For chassis, there are a lot of different options as long as you are careful with airflow/heating problem and using a stable and sufficient power supply.

Since these RAID are mostly are being used for serious business instead of high school projects, you take the risk when building a DIY system. Being in this industry (I am a tech at Areca), I've seen people scream/cry/yell/sue...do watever when their entire work for their feature simply "dissapeared"... When such event happens, even Santa can't help you if the original manufacture/vendor simply replys with they are very sorry. Here at Areca, only the most experience RAID data recovery techs are allow to touch failed system because they are the only one with the slightlest chance to perform a successful data recovery under so many different variables. I am not trying to bash people building DIY RAID but a lot of testing and engineering design cocepts are essential when combining so many different hardwares together. After all, you use your best judgement and do not just do something after reading just one article.

Happy RAIDing. :cheers2:
Click to expand...


Good scoop!
 
James Miller said:
Long time lurker, first time poster here. Just thought I'd give my 2cents worth on the DIY RAID thing. After much pain in the past (mostly on other people's systems!) using "Fake" RAID cards (ie non hardware RAID) I went with an ARECA 1221X ($860 Au with shipping) and a Sans Digital 8 bay tower (http://www.sansdigital.com/towerraid/tr8x.html) (I paid around $500 AU)with 8 1TB Hitachi drives (about $110 each AU when I bought them, even cheaper now). Running RAID 6 I get about 440 MBread/400 MB write per sec with the array 80 percent full.

I get ridiculously higher speeds with the array empty, but does anyone have their RAID empty? I've tried RAID 0 using the internal bays of my Mac Pro, but the overheads and risks just aren't worth it. That way lies pain. Trust me, I've been there. Clients don't care why the data went away, they just care that it did. $2300 AU is not a lot for that level of storage,
if you live in the US, you'd get it a fair bit cheaper cos you don't have to pay the shipping fees I do (total shipping for all the bits was around $350 AU)
and you have more choice of suppliers also.

Let's be honest, given that RAID stands for Redundant Array of Inexpensive Disks and there's no redundancy whatsoever in RAID 0, shouldn't we call it "AID 0". Coincidentally, 0 AID is exactly what you can expect if you lose data on on of these setups. If you're happy to back up to other external drives from your AID 0, that's great, but isn't that exactly what RAID 5 and 6 do all the time as a matter of course?

I don't even fully trust my RAID 6, I back everything up from it at the end of every days work using Sync Pro to another less expensive RAID 5 box via ESata. Yes, I could get higher numbers if I used RAID 5, but I prefer the comfort of knowing my files are slightly safer.

On the subject of speeds, I used the Blackmagic disk utility, I can't help but notice noone quoting numbers is mentioning where those numbers are coming from, which makes comparisons pretty useless really. Xbench for example gives me much higher numbers for the same array, but I don't think they are all that accurate. Because I'm a nerd, when I first built this array, I spent a few days testing.

One of the things I did was to play uncompressed 10 bit HD out
to my monitor (HP dreamcolor via Decklink with an AJA convertor so that I can actually use the Dreamcolor's engine) and physically ripped a drive out whilst doing so. Not a hitch. Ripped a second drive out and it kept right on playing. Whilst I wouldn't advise doing this as a daily exercise, it does give me an awful lot of confidence in my setup, that I doubt I would have in any of the non hardware RAID solutions.

Given the price of these things has plummeted in the last 2 years, I can't understand why anyone who depends on fast reliable storage would settle for anything less. My setup is a very inexpensive solution for what it is, I looked at all the options available and there was not a lot of difference in price between hardware RAID and port multipliers. Even RAID 0 port multiplier solutions are not that much cheaper, in fact way more expensive when you factor in the inevitable downtime when a drive does die and you need to recreate all the data you lost. Really, we're posting on a forum where we're talking about dealing with footage from a pretty expensive camera (I know, I know, its cheap compared to the other big guns, but hey, I was using HDV and DVCPro before this!) and we're quibbling over a few hundred bucks, maybe a grand. Personally, Once I've acquired footage, I don't really feel the burning urge to go shoot it again......... Your mileage may vary.
Click to expand...


Not everyone needs parity and redundant power supplies.
Don't pay for more than you need.
 
Last edited:
I built a DIY RAID-6 array, 6TB usable, using an Areca 1221X card, EnhanceBOX E8 ML, 2 miniSAS to infiniband cables and 8x1TB Seagate drives.

I've had it running for almost a year and other than having to upgrade all of the drives' firmware right out of the box (this was during that "bricking" phase that Seagate went through - did nothing to inspire confidence in their product...) it's been flawless.

Total cost under $2400 Canadian one year ago.
 
Some day, we'll get a concrete DIY RAID configuration that can be matched up to the much Ballyhooed CalDigit reputation.
 
Bob Gruen said:
When you step back and look at the amount of data that we are working with what you find is that there is no magic bullet as a storage solution because we all have different business dynamics.
Click to expand...

OK. I plan to put out three workflow/business model parameters for people to respond to. Your point is well taken, though I'm guessing that 90% of users of RAID solutions on this forum are single workstation clients that want a RAID 5 or 6 for uncompressed work and not teams withing to network to a shared array.
I am completely surprised that RAID 0 even exists as it basically says "You 100% believe that no drives will ever fail at the moment just before you execute a manual or automatic backup". That's just crazy.

So why don't we start by asking:
What is your workflow/business model?
 
James Miller said:
Let's be honest, given that RAID stands for Redundant Array of Inexpensive Disks and there's no redundancy whatsoever in RAID 0, shouldn't we call it "AID 0".
Click to expand...


Redundant Array of INDEPENDENT Disks = RAID

I've heard the inexpensive thing so often I almost believe it! Totally agree with your post though, Just had to be the correctness police on that one :)

RAID 0 = Laugh Now / Cry Later
 
People are perhaps being a little too hard on RAID 0 setups in this thread. Yes, if you trust your only copy of important data to a RAID 0, you're asking for disaster. But if you trust your only copy of your data to a RAID 5, you're also asking for disaster. This is especially true for DIY desktop RAID 5 configs not put together by storage experts and often lacking battery-backed cache. Accounting for all failure modes, I wouldn't be too surprised if DIY RAID 5 setups lost data more often than individual hard drives; there's theoretically more redundancy, but there's also more complexity, and therefore more that can go wrong.

The key point here is that RAID -- any level -- is not a backup strategy.

With an understanding of its limitations, RAID 0 can be a useful part of low-budget post workflow. Just don't store any files on it that you aren't prepared to lose at any second. This is not especially hard, since you'd generally be using your fast RAID to store footage files and render files. The former don't change much, so it's not too hard to make sure you always have fully up to date copies on safer storage, and the later can be regenerated.

Of course if you have an environment where taking the time to re-render or re-copy files in the event of a failure would be unacceptable, then RAID 1 or RAID 5 are obviously better choices. Given that it doesn't substitute for backups, this is what redundant RAID is really about: availability.

Incidentally, what drives me really nuts is people using G-RAID external drives in RAID 0 configurations for on-set footage downloading, which in my experience is extremely common. As compared with single-drive externals, of course, these drives are heavier, physically larger, and double your chances of data loss. Great choice for location shooting, right? Of course they are faster, but Red's current mags (and FW800, which most people still use for downloading) aren't fast enough for it to make much difference in download times.
 
It's been my experience over the years, and especially lately with SSD tech, that disks used in RAID configurations fail faster.

That said, I proudly run RAID-0 configurations where they make sense. Never put your only copy of anything onto a RAID-0, that is just asking for trouble. But often times, the performance and simplicity of a RAID-0 are a huge benefit.

My own experiences have also shown me that RAID-5 is rarely more secure. RAID-5 is fine and dandy and protects you if one of the drives has a failure. But what if the PSU in the RAID cabinet blows up and takes 2 or more drives with it? Sorry, but I find a lot of amusement with people talking down at RAID-0 and then praising RAID-5 and RAID-6 as if they're the ultimate in security.

Chirs, you are totally correct. "The key point here is that RAID -- any level -- is not a backup strategy." Oh, how true that is.
 
You must log in or register to reply here.
Back
Top