Building a storage PC

[Elyone]

EDIT: Please see post below that says UPDATE for updated idea!





I am trying to plan a data storage PC for my lab.
Of course I would prefer to buy a PowerVault type of server, but atm we cannot afford that. And we need more storage.

So I was thinking of building a full tower PC, with something like 6 x 2TB drives in RAID 5 + an SSD for OS. This will give me about 9TB net space.
Problem is SATA III spaces.
I cannot seem find a mobo that will support 6 SATA III drives.

I did find this: http://www.asus.com/ROG/MAXIMUS_V_EXTRE ... ifications that seems to have a lot of SATA connections.
But I am not sure if I can setup those 4x SATAIII under something called "ASMedia® PCIe SATA" controller (???) , together with the other "2 x SATA 6Gb/s port(s), red," in RAID 5.
the board also has + 3 x SATA 3Gb/s port(s), black + a mSATA I can use for the SSD.

My questions:

1. Can I use the 4 + 2 SATA spots in a single RAID configuration? Or only either the 4 or the 2?
2. Is this RAID onboard? Seems so. (i.e. Do I need a separate RAID controller?)
3. Any other ideas for a mobo that can handle 6 SATA III + an OS drive? Can be less expensive of course, this is more of a gaming board - a bit overkill.
4. Any reason why not to use 3TB drives? If I can only use the 4x SATAIII in RAID, this might solve it.

Thanks,
el

[vbironchef]

http://www.legitreviews.com/article/2092/1/. This might help.

[egloeckle]

6 x 2TB drives in RAID 5

You are making me twitchy...

Start here:
http://forums.legitreviews.com/about43565.html#p217308

Then try:
http://www.zdnet.com/blog/storage/why-r ... n-2009/162
Ignore the 2009 part, pretty sure the author was just trying to drive the point home a bit more directly with that. Do a google search for raid 5 is dead and there are plenty of other links with the math on why its a bad idea these days.

I have a feeling several people are going to come out and say how RAID-5 is fine and how home and small businesses are different (they aren't). I dont want to get into this again, just do your research and decide if you can accept that level of risk.

Also, look at a Synology unit. They do just about everything you could want out of the box and will end up costing about the same as building a machine. They support NFS, SMB, AFP, iSCSI, and tons of other protocols which covers most use cases and the setup/management is dead simple. There are other brands as well, though personally we have had the best luck with the Synology units (mostly iSCSI/VMware setups though so YMMV).

[Elyone]

Thanks for the replies.
1. I already intended to use the Red disks.
2. Wow. I always assumed RAID 5 is still standard. After your links, I started reading up and there is lots of hate against 5. So i am now thinking of RAID 10. This will certainly cost us space though.

So, lets say i go RAID10.
After further research, I found this AMD chip based mobo SABERTOOTH 990FX (http://www.asus.com/Motherboard/SABERTO ... ifications), that has 6 SATAIII spots for RAID. So:

What about 6x 3TB drives in RAID 10? Does this sound good? Then I can use one of the 3GB/sec spots for the OS disk.

thanks
el

[KnightRid]

All depends on how much redundancy you want and if you have a separate unit for backing everything up.

Yes, I am one of the RAID5 is just fine crowd as long as your data is already backed up or you dont really need any of it backed up. Losing 3TB of space just to have another parity drive in RAID 6 is a waste of money and space for me and anything that takes even more space away is the same....for me.

If you are using it for a backup device then you need massive redundancy. If you have the budget to be able to afford a lot of extra hard drives or add-on units, then by all means build it for redundancy!

Here is a great chart - http://www.adaptec.com/en-us/_common/co ... par_wp.htm - from adaptec.

I would also suggest a standalone NAS unit as motherboard RAID is nice but if that motherboard fails you may have problems ever reading your array again on a different system. If you were planning on a RAID card in the system, compare costs of the unit I just bought Synology DS1512+ (I put 5 3TB WD RED drives in it) and you will see how the NAS unit will easily beat the computer now a days (yes RAID is still overpriced either way). Plus with the unit I listed you can add on up to 2 more 5 drive units to expand to 15 drives total so you can have the super redundancy and still have space to work with. Heck do a RAID 1 and have everything backed up if you want as you should have plenty of space as long as you have the money for the drives.

Also look at the power requirements of a computer compared to a NAS. BIG difference usually.

[egloeckle]

If you are using it for a backup device then you need massive redundancy. If you have the budget to be able to afford a lot of extra hard drives or add-on units, then by all means build it for redundancy!

Eh? Why do you need more redundancy for a backup device then a live storage one that typically has a higher workload? This just makes no sense at all, you would rather have more redundancy for live data then the backups in most situations, and pretty much every situation that would affect the home/SMB segment. Ideally backups in that situation are only to recover from mistakenly deleted files or to roll back on files for various reasons. You would rather not have the downtime associated with things going down and having to recover from backup in the first place.

As for the OP, for 6 drives, I would just run a RAID-6 array. You wont lose as much space and to be honest, the performance hit wont affect overall performance as the bottleneck will be your connectivity (unless you are planning bonded nic's or a 10gbe setup). The performance difference between RAID 5 and 6 is one drive. The parity penalty is the same, you just write out that block to 2 drives instead of one. Maybe a better to understand way to put it is a 6 disk raid 6 array should have identical performance to a 5 drive raid 5 array. With that many disks, even counting the parity overhead you should still easily overwhelm a single 1gb (125MB/s - overhead) connection.

I can give you a more technical breakdown if you want, but for now ill avoid that as it would become a bit rough to read.

Edit: while typing this I ran a simple test on my home setup (Synology DS1512+ with 5x 2TB WD Ent. SATA drives in raid6 connected via a managed switch to an HP ML350 running VMware ESXi 5.1. OS: Win2012) and a single NIC will still very easily be the bottleneck by quite a bit. Unless you are planning on implementing bonded nics I think raid 10 is going to be a waste in your case. Ill admit my setup is a bit larger then the typical home setup, but I think it shows my point about the nic being the bottleneck.

As a side note on my edit, I just want to state how much Win2012 makes me want to kick puppies. That interface on something remotely accessed is just infuriating.

[Elyone]

I actually WANTED to run RAID 6. While I am decently technically proficient at hardware(I built 3 systems already, but none with RAID) I am not an expert.

According to the mobo manual, it only has 1,5, 10.

But now I am further confused, after what Knight said about not being dependent on the mobo for the RAID.

So, should I use a separate controller? (recommendations?) But can't that ALSO go kaput?



I need the system for mass storage of image data, we create about 40GB a day.
My workflow is:
- acquire data on local computer attached to microscope.
-transfer to the storage
- analyze on other workstations off the storage

atm, we do have 1 "real" SAS based NAS - a PowerVault - but we only have 2.5 TB net (we share it). We can add another drawer to it, but right now we cannot afford it until next year when another grant comes in - and we will, but for now I can only spend less. So I had come up with this idea.

[KnightRid]

I still suggest using a separate NAS device. Building a computer for storage will require a RAID card if you want to be able to move the drives to another system at some point and that will add even more cost to the system. I found this out when I was thinking of building a RAID system on my computer. For some reason if you build it using the onboard MB connections the RAID is a pain to hook up anywhere else and actually see the data, and in most cases, impossible.

Grab a 1512+ for $800 add 5 drives of whatever size you want and by all means, run RAID 6 if you prefer! Problem is solved and you will be happier than using a computer for the same purpose.

As Eglo said, you may find a bottleneck in your gigabit network if your router doesnt bind 2 channels together (the 1512+ has 2 gigabit ports that can be bound for, theoretically, double the speed. Now if you dont move a whole lot of data at the same time, you wont notice it.

[egloeckle]

I still suggest using a separate NAS device.

Grab a 1512+ for $800 add 5 drives of whatever size you want and by all means, run RAID 6 if you prefer! Problem is solved and you will be happier than using a computer for the same purpose.

Exactly what I would/was recommend(ing). If for some reason the Synology 1512+ doesnt meet your needs, use another vendor's BYOD NAS solution that does. You will get an almost ready to go system with more functionality than you can realistically build on your own (without excessive amounts of headache). Though I can stand behind Synology's units, having used various units from the 2 disk ds212 (i think) up to some much larger RS34xx (dont remember the exact model and too lazy to log in and look right now) they are very solid units.

[Elyone]

UPDATE

After reading extensively on all the info here (thx guys), I came up with a modified idea.

1. I still don't want a local NAS, rather a full PC with lots of HDDs. I did look into it and the NASs I would be able to purchase within the confines of our institution are too expensive. However, they ARE letting me custom build my own system, because "you are not buying a completed item so you don't need to use the venders we have agreements with".
2. I am scared now of RAID the way I wanted to do it using the onboard mobo controller. Too risky.

SO:
I was thinking of still going 6x 3TB, and simply doing an automated sync with software every day at the end of the day. No RAID. No need to worry about controllers. All of my data will be stored on 2 HDDs.
Of course, I can always lose a single day of data, but the way we work is that we do not delete the original data from the acquisition computer for a few days anyway.

in other words, this is a hack of a solution, but one that seems that would work for us. OK, so, NOW what can go wrong?

[KnightRid]

UPDATE

SO:
I was thinking of still going 6x 3TB, and simply doing an automated sync with software every day at the end of the day. No RAID. No need to worry about controllers. All of my data will be stored on 2 HDDs.
Of course, I can always lose a single day of data, but the way we work is that we do not delete the original data from the acquisition computer for a few days anyway.

in other words, this is a hack of a solution, but one that seems that would work for us. OK, so, NOW what can go wrong?

So how are you going to tie all the data to 6 hard drives but use only 2? You really confused me with that one.

You would have to set each machine up to send data to specific drives and not 1 RAID array. So machine 1 may use drive G, Machine 2 Drive H, etc. with the data then mirrored onto another drive..?? This is really patchwork and whatever moron told you that you cant buy a NAS should be fired or at least written up for costing the company way more money than a simple solution would. That NAS we mentioned above should not go above the cost of an entire computer.

If vender agreements lock you in then tell those vendors you want the same pricing you can get from other places or you wont renew those lame agreements!

If the only option i had was to build a computer to house storage, I would just buy a blu-ray burner and some discs to backup data that is taking up room on my existing raid setup until I had the money to buy another one. I would not build an entire computer to do the work of a NAS and not even have the RAID abilities for the same price or more than the NAS would cost.

Just my 2 cents but I would tell them to shove it and if they wont approve the NAS then work will stop or data will be lost.

This has to be a government facility....seriously, how stupid are the people running this place?

[Elyone]

Ouch i meant of course 3 disks.
i.e. 3 will be storage, and 3 will be redundant at 1:1. So the computer will be connected to the network, the students will have access to say disks d,e,f and at night I will back up incrementally to g,h,i.
I know it's patchwork.

You seem to be comfortable with RAID. But it seems like so much can go wrong with that too. I read some horror stories about the Synology and similar types of NAS. People cant access the data because the controller died. Does not matter if the controler is in a black box, or a computer case, iIf it broke, I lose data, no?
Just seems to me that a computer, while having overall less performance, at the end is more robust and I have more control over it. A bit more hassle, but this way I can yank out a drive and stick it in another computer within minutes and have access to my data again. We make so much, that just keeping track of disks would be a big hassle.
I will rethink again though, thanks.


And they ARE actually pretty stupid. They refuse to let the main IT computing center give data storage service to the academics, only to the administrative. I would love to simply pay for more storage over the network. This is the "brainchild" of the General Manager, whom was never an academic. clueless. if we buy our OWN PowerVault like NAS, they will install it on their center, but they cant give us space that they organize.