RAID

How much storage do you need? How fast do you need to move data?

>>62348670
probably around 4TB should be enough
Just for storing it it doesn't need to be exceedingly fast, but it's good to have speed for editing

>>62348685
Just get RAID1 with a couple buffering SSD. Keeps it available well enough and doesent cost a fortune. Remember that a raid IS NOT a backup, if you care about your data survivability get a backup plan.

>>62348658
Get a 2 bay NAS. Get two 4TB drives and do a raid 1. If you lose a disk you still ok.

>>62348658
RAID 5, 6 or 10 is the best

>>62348765
>>62348814
I would suggest RAID 10 for increased speed.
>>62348818
>5,6
>best
Maybe with SSD's, rebuild times with current sizes is simply not feasible

>>62348818
RAID 5 and 6 are gimmicks with today's HDD sizes.
RAID 10 with 2 mirrors is the only way to be safe but it also wastes disks.

>>62348850
>>62348898
Why do you care about rebuild times?
If a drives goes wrong switch it and tell the server to rebuild, it's running anyway

>>62348658
Software RAID is best RAID.
Other than that it's about buying extra disk space for improved redundancy.

I assume you absolutely can't afford loving footage?
If so you'll want at least one backup in addition to the redundant array.
Personally I would even go as far as having two backups, so you can erase one to re-write it and still have the second backup in case something happens in the meantime.

So for 4TB you could buy 4 4TB drives: 2 for RAID 1 and 2 for backups.
Alternatively you could have 3 2TB drives in RAID 5 (plus 2 4TB drives for backup) but I'm not sure that's cheaper and RAID 5 is far more failure prone than RAID 1.

>>62348924
Not him but he did touch an issue.

With RAID 5, a drive can be dropped if it doesn't respond in time.
This has happened to me numerous times.
After which rebuilding took about 30 hours.
That's 30 hours of heavy disk usage where you pray no other disks get dropped.

With RAID 1 if something like this happens you still have 2 working copies.

why are naas devices so bullshit in price? for the price of a 6 nay nas i can build my own pc and just use the sata ports for raid

>>62348924
That's the problem you retard.
Since HDDs have a rather limited write speeds (~125 MB/s for most consumer models), the rebuilds take a long time for drives larger than 1 TB.
During that time, all other drives are 100% utilized and there's a high probability of another drive failure or an unrecoverable read error.
On RAID 5, you are instantly fucked. RAID 6 is a bit more resilient at the price of more complicated rebuild and the fact that it will take even longer once a second HDD dies.

>>62349070
That's because you were using drives without TLER, which aren't meant to be used in RAID setups.

>>62348898
>RAID 10 with 2 mirrors is the only way to be safe

It's not as safe as RAID 1.

You use RAID 10 for speed or when your disks aren't big enough.
Otherwise, just stick to RAID 1.

>>62349111
That is true.

But the magical "RAID ready" drives (identical drives just with a different value for the max timeout set in the firmware) were more than double the price.
Which made RAID 1 cheaper.

Not sure if (((hard drive manufacturers))) still do this?

>>62349116
No, RAID 10 is superior.
You get better read and write speeds because of striping and rebuild is a simple mirroring of one disk.
And it scales dramatically better because you can stripe data across many small disks that rebuild quickly, while with RAID 1 you need huge disks that also share the rebuild weakness with RAID 5 and 6.

>>62349095
That´s the price for stupidity

>>62348658
Why would you want to "get into RAID"? RAID should be a tool, not a goal. What are you trying to achieve?

>>62349278
For unexperienced people setting up and running a good nas can be an achievement. Its a good idea to ask before burning 100s to 1000s on it

>>62349199
Not him, I agree with most of what you said.
RAID 1 has better rebuild than 5, since it's simply a copy of your healthy drive. RAID 5 needs to calculate the parity on every block.
6 is even worse, since you need to calculate TWO different parity bits.

But yeah, you're right.

>>62349362
Yeah, but the conceptualization of the question is bad. If you start out with "I want to do RAID" then you have found a solution looking for a problem. Such a quest should start with a problem that needs solving, not a shiny tool that needs applying to things.

>>62349225
i get that its a computer and all but theyre not modular at all and cant do anything but nas

https://au.pcpartpicker.com/list/7TFb3F

cheaper than every 8-bay nas on the market and has way more room for upgrades and usage

>>62349390
>RAID 5 needs to calculate the parity on every block.

Which with modern CPU's takes hardly any effort at all.

For both RAID 1 and RAID 5 the rebuild will be limited by the write speed of the new/broken disk.
Rebuilding RAID 5 is arguably easier because it puts less strain on the remaining disks (so you don't notice it as much as you keep working).

>>62349564
>Which with modern CPU's takes hardly any effort at all.
With RAID 1 there's no CPU involvement.

Rebuilding RAID 5 is arguably easier because it puts less strain on the remaining disks (so you don't notice it as much as you keep working)
Nope, you're wrong.
It needs to read from ALL disks to calculate the parity and rebuild the disk. And then it needs to serve your requests There's no way this will not be noticeable.

I once inherited a RAID 5 fileserver from a previous sysadmin. No one from the sysadmin team listened to me when I said to move that shit to RAID 10 before a disk dies.
Needless to say, a disk died a year down the road and it nearly started a riot in the company, simply because it was too slow while rebuilding.

>>62348658
>>62348685
Raid 5, 5x1TB 7200 RPM SATA III drives.

>>62349070
>>62349107
Irrelevant. You aren't supposed to have a failing HDD every other day, quite the contrary. Also, ***RAID IS NOT A BACKUP SOLUTION*** you fucking faggots.

>>62349745
>You aren't supposed to have a failing HDD every other day

No, but with RAID 5 the likelihood is there. - that's the point.

>>62349745
raid 5 is too dangerous. what happens if a disk fails during the rebuild (say two of the original disks were faulty and from the same batch)? youre fucked

>>62349803
>youre fucked
You'd use your backups, since raid is not one. On the availability side though, 6 is always better

>>62349918
10 offers a legitimate backup solution, especially with large data that cant reliably be stored off site or in the "cloud"

>>62349745
>You aren't supposed to have a failing HDD every other day
That's not the point you retard.

>Also, ***RAID IS NOT A BACKUP SOLUTION*** you fucking faggots.
So? None of the posts you respond to allege that.
The point of RAID is availability first and foremost, which goes down the drain when:
1. all disks are 100% utilized because parity calculation for rebuild, or
2. the whole array fails.

But I feel like all this is lost on you, because you can still cover your ears and make up another strawman.
Fucking nu/g/ I swear to God.

>>62349803
1) never use two disks from the same batch; they better have some properties but this doesn't mean they should come from the same batch
2) monitor your disks consistently and predict failures
3) ***RAID IS NOT A BACKUP SOLUTION***
4) rebuilding in the worst case scenario is already something _beyond_ OP's goal

>>62349780
and when and if it happens, you may have a chance to backup something you sorely missed after 400 gazillion hours of rebuild time, then replace every disk of that batch and/or any near-to-failure disk.

>>62349979
You're a fucking idiot.

File: 1485455469191.gif (205KB, 160x90px) Image search: [Google]

205KB, 160x90px

>>62349960
>in the "cloud"

File: 1502010599212.jpg (58KB, 430x760px) Image search: [Google]

58KB, 430x760px

>>62350009
>You're a fucking idiot.
Great arguments there you autist.

>>62350137
coming from the guy who says raid 5 is good

>>62350180
meh, the URE failure thing has been hyped up beyond what its actual occurrence rate is.

You can get NAS specific HDs that have a URE error rate of 1 bit out of 125 terabytes.

Not impossible to happen, but very much away from the "50%" failure rate that has been throw around everywhere.

File: 1461359323014.jpg (18KB, 201x201px) Image search: [Google]

18KB, 201x201px

>>62350180
>projecting this hard

>>62350245
red or gold?