RAID5 on a new microserver?

[jackvdbuk]

hi guys, ive not really dabbled in RAID but i currently have a NAS RAID0 setup and am really worried of losing all the data on there.

ive just bought a hp microserver and was thinking of finally getting round to creating an array that includes redudancy and so at first always thought of going raid1 but then the cost incurred of getting another 2 2TB drives is rather... well expensive!

so brings me to RAID5 (which seems to have to be software based) and that way only need one more drive... much better!

the server will mainly be used for streaming/downloading e.t.c

any flaws in my logic? preferably own experiences

thanks

[jim]

This is the main problem: http://www.zdnet.com/blog/storage/wh...ng-in-2009/162

All depends on the size of your array, but generally speaking the larger the array, the less valuable RAID5 is. That's why I'm running RAID6 on my Microserver, but obviously that means another redundant disk.

Also you need to decide what OS you're going to use - the Microserver only support RAID0 and RAID1 in hardware, so any RAID5 or 6 solution will need to be done in software. I think most of us are using Openfiler but it's not the easiest piece of software to configure in my experience.

How many drives are you intending to use, and what capacity are they?

[shaithis]

Firstly, getting off of RAID0 is a good move. It is good in certain situations but yours is definately not one of them!

RAID5 will give you redundancy. You will take a small hit on write performance but for what your planning on using it for, that shouldn't really matter.

I'm sure you probably know this (or have read it before) but I really cannot enforce enough the moto "RAID is not a substitute for backups".

[shaithis]

This is the main problem: http://www.zdnet.com/blog/storage/wh...ng-in-2009/162

I've seen that posted many times.....great bit of theory but ultimately, something I have yet to see and I've been using 4TB+ arrays for years both at home and at work.

[jim]

I've seen that posted many times.....great bit of theory but ultimately, something I have yet to see and I've been using 4TB+ arrays for years both at home and at work.

It's all a probability game though, isn't it. If the manufacturers are being conservative with their specs, then you're right and we don't need to worry. If on the other hand they're right, then you could easily lose data.

Point is, you're probably right, and it'll be fine, but one thing's for sure - the bigger your array gets, the riskier RAID5 becomes. Even if the risk is slight, it's still increasing. That's why I'm sticking to RAID6 from here onwards.

[b0redom]

....and don't forget the recalc times on such a weak processor leave you at a significant risk. I'm RAID-Z2 on my Microserver (similar to RAID-6) too.

[jackvdbuk]

This is the main problem: http://www.zdnet.com/blog/storage/wh...ng-in-2009/162

All depends on the size of your array, but generally speaking the larger the array, the less valuable RAID5 is. That's why I'm running RAID6 on my Microserver, but obviously that means another redundant disk.

Also you need to decide what OS you're going to use - the Microserver only support RAID0 and RAID1 in hardware, so any RAID5 or 6 solution will need to be done in software. I think most of us are using Openfiler but it's not the easiest piece of software to configure in my experience.

How many drives are you intending to use, and what capacity are they?

i havent even got round to deciding what OS and such, but i was going to have a play with 2008 r2, linux, and freenas server or whatever its called to see what i like best ... never touched a server side of stuff so all new to me

im currently bordering 2.5/3tb of space used so 4tb server usable space is what im after, ill have a check of raid6 but seems ill have to buy two drives for redudancy... so much money!!

also what other than rebuild times are problems with software raid? edit: sorry i was being lazy, google has been a good help!

thanks so much you guys

[jim]

i havent even got round to deciding what OS and such, but i was going to have a play with 2008 r2, linux, and freenas server or whatever its called to see what i like best ... never touched a server side of stuff so all new to me

im currently bordering 2.5/3tb of space used so 4tb server usable space is what im after, ill have a check of raid6 but seems ill have to buy two drives for redudancy... so much money!!

also what other than rebuild times are problems with software raid? edit: sorry i was being lazy, google has been a good help!

thanks so much you guys

Ah you'll be fine with that, it's when you get up to 10TB-ish that RAID-6 becomes more of a necessity (although as you can see, not everyone agrees on that).

I would advise sticking with FreeNAS or OpenFiler to be honest, this is exactly the sort of thing that it's designed for. Software RAID isn't ideal, speed is the big problem, but when we're talking about fairly simple arrays like this it's not a major issue. If you want bleeding-edge performance, then you'd be crazy to use software RAID when hardware is an option, but for a home NAS it's not going to hurt you.

Just make sure that you've got sufficient blank disks, and have a play around. You'll want to try out a few different configurations so don't try to do anything with live data until you're 100% sure that you're happy with the server config. You've got a rough idea of what you want to do, so just have a play around and see what suits you best, then go for it.

[jackvdbuk]

ahh cheers, it seems as long as its mainly read only operations, raid5 is perfect and much cheaper!

just for future knowledge, what sort of server operations would warrant the higher read/write speeds raid 10 or 6 brings? would it be things like database, webservers, and such?

[GoNz0]

i would not use raid5 as a backup solution, i use raid 5 to mirror whats on the PC & weekly backups on it.

I did have all my films stored on my Raid5 NAS until it dropped a drive then dropped a 2nd one mid rebuild.

i then got a 2tb storage drive for the PC so i have a backup of a backup

[jim]

ahh cheers, it seems as long as its mainly read only operations, raid5 is perfect and much cheaper!

just for future knowledge, what sort of server operations would warrant the higher read/write speeds raid 10 or 6 brings? would it be things like database, webservers, and such?

I would think RAID6 is slower, since it's adding a second parity drive, but I could be wrong. Yeah, that would be typical RAID10, I know on HP servers RAID10 is the default configuration (or at least it was). Even file storage can be RAID10, all depends on how many clients you have trying to access the files simultaneously. When you've got fifty machines trying to get different files at the same time... you can imagine.

And as Gonzo said, I would always make sure that whatever you have on your RAID5 array you have a backup of.

At present, I have a RAID6 server, and then two 3TB drives that I plug in once a month, backup to, and then put on the shelf. However good the RAID level is, you can't guarantee your data won't be vapourised.

[GoNz0]

well tbh my raid5 NAS box is just a backup of my pc including a weekly windows drive image! I would never trust it to hold important stuff i cannot replace such as pics of my little girl.

losing all my films after just using the NAS box to store then was the wake up call i needed to add a 2tb drive to the PC so i could have a true backup on the NAS.

really important stuffs mirrored on my rack server and sky drive.

so baby pics are backed up, and backed up and backed up and backed up (and backed up again on my laptop)

i even take the NAS drives to dads when going on holiday

[Biscuit]

I had loads of problems with getting permissions to work in a multi-user environment with FreeNAS. Openmediavault was a lot easier in that regards however its user base and community is still small so dont expect much support!

[mikerr]

i even take the NAS drives to dads when going on holiday

Lol - every month I swap a backup drive with a mate;
I have his last backup drive at my house, mine at his.

Off-site backups... in case of burglary/fire etc

[DanceswithUnix]

It's all a probability game though, isn't it. If the manufacturers are being conservative with their specs, then you're right and we don't need to worry. If on the other hand they're right, then you could easily lose data.

Point is, you're probably right, and it'll be fine, but one thing's for sure - the bigger your array gets, the riskier RAID5 becomes. Even if the risk is slight, it's still increasing. That's why I'm sticking to RAID6 from here onwards.

It is a numbers game, though I don't think the numbers are changing that fast.

As drives get bigger the amount of forward error correction added to avoid bad sector reads increases. This was the main driver of the change of big drives to 4K sectors internally as the FEC overhead on 512KB sectors was getting too much.

On top of that, buying enterprise class drives costs a little more and gets a magnitude better MTBF.

Just avoid buying all your raid 5 drives at the same time, so they aren't all from the same batch, and hence fail at the same time

[drsmithy]

....and don't forget the recalc times on such a weak processor leave you at a significant risk. I'm RAID-Z2 on my Microserver (similar to RAID-6) too.

This myth has to die. CPU power hasn't been relevant to software RAID performance for a decade. The CPU in the N40L can calculate parity at ca. 4GB/sec (in Linux). It's not a bottleneck.

For reference, my N40L will do a "zpool scrub" (roughly equivalent to a RAID rebuild, though much more CPU intensive) at around 100MB/sec per drive. Which is about all the old Spinpoint F1s have to give.

FWIW, I wouldn't use RAID5/RAIDZ on drives bigger than 1TB - RAID6/RAIDZ2 only. This is seen in the real storage industry as well - EMC's best practice is RAID6 for their 2TB drives. NetApp's RAID-DP is similar to RAID6, but they've used it for all kinds of disks for years.