OK today I received my G-RAID 8 TB drive. I am new to RAID and was wondering what is the best practical setup. It comes configured for RAID 0 which offers the best performance. However if any of the drives fail I lose ALL data and that scares the crap out of me. I was thinking that RAID 5 might be a good option but Apple's Disk Utility app does not allow this. It doesn't allow any RAID numbers, actually, It just has labels like "mirrored" and "striped". I did not seem to be able to change anything with how the RAID is set up. How would I do so? And what RAID would you guys recommend? Would RAID 5 shrink my 8TB down to 2TB? I'm not sure if 4 or 2 drives are enclosed.
Striped is Raid 0, Mirrored is RAID 1. For 4 or fewer drives, RAID 1 is good. BUT, verify that your enclosure supports swapping a dead drive for the controller card to rebuild it. Not all enclosures support that function. Otherwise, Mirrored in Disk Utility will work just fine for you. An array that small, RAID 5 would be overkill.
OK, you have to set the drives to create the RAID with, then click "Create". But there needs to be a set of drives listed first. Basically, you have to break the RAID apart, and rebuild it. G-Tech should have formatting software of their own that probably is more detailed and works better than Disk Utility.
I think the thing to keep in mind is the benefit of having a G-Raid t-bolt drive is speed which you will only get if you leave it RAID-0.
The other thing to keep in mind is that fast raid storage is really good for one thing, spitting footage to your NLE as fast as possible. RAIDs are like a sports cars, they're not meant to be safe and reliable, they're meant to be fast.
(An exception is a raid built for backups with redundancy but the G-Raid is a sports car)
The reason on other posts I have cautioned against RAIDS is because I have had clients who's G-Raids have died and it was their only copy. The risk is higher on a RAID-0, but the benefits are huge when cutting.
To offset the risk, I back footage up to cheap USB3 backup drives. I also keep backups of projects (now libraries) on different drives. If I'm really paranoid, I Dropbox them just to be really safe. If my raid array craps out, I can still get to my work. Yes it will be much slower, but I won't lose anything.
The RAID-0 on that drive is worth it as long as you have your backups elsewhere.
With spinning disks, I doubt you'll see much difference between RAID 0 and RAID 1 or 5, seriously, with T-bolt. I'm using a T-bolt 2 RAID 5 and it is plenty fast for multicam 4K. RAID is not a backup, and should never be relied on as a backup. The redundancy is only one level of safety, but is not replacement for a backup. I don't think anyone has ever advocated using a RAID without a backup system.
BenB wrote: With spinning disks, I doubt you'll see much difference between RAID 0 and RAID 1 or 5, seriously, with T-bolt. I'm using a T-bolt 2 RAID 5 and it is plenty fast for multicam 4K. RAID is not a backup, and should never be relied on as a backup. The redundancy is only one level of safety, but is not replacement for a backup. I don't think anyone has ever advocated using a RAID without a backup system.
Aside from the spinning drives part I am in complete agreement.
The G-Raids in RAID-0 get about 315MB/s read speed. The same G-Drive generation gets about 160MB/s-ish read. Doubling the drives doubles horsepower in a way.
Check out the read write specs on their website, they are (at least when the drive is brand new) pretty accurate. My trusty eSATA 6TB G-RAID gets about 250MB/s when freshly formatted.
the G-Tech 8TB ThunderBolt drive you bought contains 2X4GB drives, most likely the Hitachi 7K4000. G-Tech and HGST are now owned by Western Digital (WD). Since it contains only 2 drives, RAID 5 is not possible as RAID 5 requires at least 3 drives. Also as far as I know, Mac OS does not support RAID5 and you can only use hardware controlled RAID 5 solutions. You could if you wanted to spend more money, purchase an identical unit, daisy chain it to create a RAID 10 (1+0) RAID. RAID 10 requires at least 4 drives.
As far as I know, the G-Tech box does not have a separate "hardware" RAID controller chip and is configured as a software RAID via Disk Utility. The G-Raid box goes TB1 4 lane controller (pass through to 2nd TB port) to SATA 3 control chip with 2 connections on the bus.
RAID 0 (striped) will give you 8TB (actually slightly less) capacity, but no data protection. 1 drive fails, all data lost. You do gain ~2X the write speed (as Darren has said around 300MB/s) over RAID 1 (mirrored). RAID 1 will only write about 150MB/s. This is due to the hard drive read/write speeds. Replace the hdds with ssds and read/write speeds would vastly increase, this would void your warranty, however.
RAID 1 will give you around 4TBs capacity and some data protection, as the computer is writing the same data to both drives at once, if one drive fails, the read/writes just continue with the good drive. However, as others have said, additional back ups are essential for safety. While RAID 1 provides redundancy, it is not the same as a back-up copy. Practice safe sex, back it up.
Given your first couple of postings, you aren't doing any real heavy editing, so in fact, you will not see any practical advantages of the TB box or RAID for that matter. Given how you described your work, you will not gain anything from thunderbolt over USB3. USB3 can handle around 4.8Gb/s. Notice this is Gigabits/second and the hard drives are measured in Megabytes/second.
The G-Tech is a ThunderBolt 1, meaning 10Gb/s port, while your new MP has TB2 (20Gb/s) ports. The G-Tech does have SATA 3 (6GBs) bus inside, but the bottleneck is the hard drives. While I think G-Tech makes nice products, I also think they are way overpriced, IMHO. Probably just started a flame war, hope not, just my opinion.
Anyway, hope this helps, Greg
"Probably just started a flame war, hope not, just my opinion."
I think you are in the wrong forums if you are worried about that!
With FCP X making new thumbnails and waveform redrawing on the fly, you just can't have enough drive speed anymore. RAID-0 will be a help in FCPX no matter what you are cutting.
An experiment I did using the cheapo Costco Seagate backup plus drives where I speed tested one with the USB3 dock it came with and the Seagate Thunderbolt Adapter. (My older MBP only had Tbolt and USB2) The thunderbolt adapter with the same drive had much better read/write speed. The USB3 connection was about 100MB/s, the tbolt adapter was 130MB/s. It's the same exact cheap green drive, still nowhere near the speed limit of USB3 yet still had about 100MB/s. On Tbolt it had 30% better speed and I ran the test five times each with both docks.
Testing aside, the real proof for me that tbolt has other voodoo under the hood is when I had to open a massive job I had moved from the raid to one of these seagate drives. Originally I plugged it in with the USB3 and it was really sluggish. Then I swapped it for the tbolt dock and although it wasn't a rocket or anything, it was so much smoother.
My limited understanding is that the chips inside each end of the thunderbolt cables take some of the heavy lifting of data processing away from the CPU which boosts performance and most important for me when editing video is stability. I'm glad they do something other than costing a lot!
So many factors are involved in transfer speeds, no matter what the interface. I think the place to begin is the type of work you are doing. If doing SD video, vs HD video, vs 4 or 5K video, single stream vs multi-cam, number of effects, codex involved all dictate what computer set-up you need.
The OP stated in his first post that he was not involved in 4K production and had minimum requirements and wished to save some $$$. He is currently using single HDDs for media/backup/boot drives. So I think that should be the goal.
If you do not require the horsepower, then as long as you meet your bandwidth/throughput needs, no faster process will improve your work.
I totally agree with you that Tbolt RAID 0 with 2 or more drives are faster, but if you are working in a format that doesn't need/use more throughput you will not realize any gains.
As to your experience in testing USB3 speeds, we just opened Pandora's box regarding technical issues. There are so many factors to what any combo of devices, firmware, hardware, drivers, software are utilized that will affect actual performance. USB3 certainly opens up lots of possibilities.
So starting from the theoretical performance of ~4.8Gb/s you have to subtract overhead. Overhead is dependent on various protocols and whether the sending and receiving device will support each protocol.
Overhead can be upward of 36% on older devices/controllers (much like you experienced) or much lower. There can be other overhead penalties as well.
You don't say what system you used to test USB3.
Early USB3 devices/controllers used older legacy BOT nonparallel routines that significantly impacted performance. Newer USB3 devices/controllers use USB Attached SCSI Protocol (UAS/UASP) that reduces overhead by up to 20-30%. They also recognize full duplex, parallel data streams on 4 channels.
How the OS, firmware, amount of free space, even quality of cable can cause different results for any given drive. So without knowing every single detail about the tested set-up it is impossible to say what to expect.
Again the question is what is required to meet the workflow.
I fully agree that Tbolt will be faster, hands down no argument. At what cost? Do you need the speed? If so go for it! If not as the OP stated, then you can save a boat load of money going USB3.
Tbolt which is based on PCIe2 also has overhead, indeed some of the same protocol error correcting as USB3 so takes a hit as well. PCIe3 avoids, really includes this overhead in the standards. It gets technical here so I won't embarrass myself by trying to explain.
Apple's implementation of Tbolt2 and USB3 is really a hack job to make it work on the older CPUs Apple is using. Essentially, the new MPs are I think are plan B until the new enterprise xeon processors are ready in 2014. The current Ivy Bridge CPUs allow more PCIe lanes, better reliability but I think unless you need the small increases in speed, you should wait for the June 2015 MP(clairvoyance). When Intel releases xeon Haswell's the MacPro will gain native USB3, more PCIe lanes(up to 40 PCIe3 lanes) and significant speed/capacity increases. Maybe even dual socket CPUs and 2 flash memory slots, non shared HDMI 4K video output.
GPU improvements will also vastly improve the overall MPs by that time.
If you are going to do 4/5/10K multi-cam/3D movies, I think this will need the next iteration of MP. If you are doing SD/HDs, even 4K now, add a second GPU to an older MP or work in Optimized/Proxy media, take the speed hit and save your $$. Mavericks really opens up some OpenCL/GL options to keep the old girl going.
Of course, major editing studios can justify upgrading now. If someone wants to give me a new MP, I'll kiss their feet and lovingly extoll it's virtues. Just been on the early adoption program too many times (since 87) to not see the benefits of second addition hardware.
Did I manage to get off topic or what?
Just my wooden nickels worth,
Hope this helps, Greg
I agree with everything from your post, you really know your stuff. I just want to add one thing which is how FCP X unlike FCP 7 can multitask.
When FCPX is generating thumbnails, resizing the timeline, rendering waveforms, etc, it's accessing all the relevant files in your event. Even though background rendering is paused while you edit, all the other tasks can gobble up the drive's throughput as they are still happening while you play the timeline.
He will still gain performance and stability benefits from the drive configured as a raid-0 even if he is just cutting 1080p prores files.
My personal experience FWIW, is that I can really tell the difference when I'm cutting off a single non-raid drive. The beachballing happens less frequently which makes me less cranky. To me anyway, the extra money is worth it.