mdogg

Hey guys,

I'm planning to make a 3TB(3x1TB) home file server.
I'm going to put all of our movies and music (lossless) on it in addition to files. I'm probably going to backup hard drive images as well once a month.

Because some of the data is really important I was also planning to backup select folders onto Amazon S3.

Disk I/O would be in short spurts

Would RAID be appropriate? Or would it be overkill...

hoth17

If the data is important, then yes I would say a raid would be nice. It would be a real time backup. I suggest a Raid 5. Here is a good explanation of the different raid setups

http://en.wikipedia.org/wiki/RAID

With a raid 5 you will not have as much hdd space, but if one of your drives dies, the raid will continue to run. Then you can replace the raid and it will rebuild itself.

__________________
For Sale:
HD-555 | SIIG Bit-Perfect Sound Card
Intel X25-M v2 80gb SSD
Complete Water Loop: Heatkiller v3 | XSPC 360 RAD | MCP655 Vario | EK 250 res
2 x GTX 280 *$225*

mdogg

Thanks for the quick response hoth17.

RAID 5 was what I was thinking of as well.

I understand RAID and all, but the way I've been taught, RAID is mainly a tool for uptime. Since uptime really isnt an issue because critical files will be on the cloud, im not sure if RAID is such a good idea.

grahamcrackuh

I'd go with raid 5, it will help slightly with uptime and it will also automatically backup your data without taking a big loss on your total size of drives.

__________________

BLinux

RAID levels that provide redundancy protect against data loss in the event of a drive failure. Backups provide a historical snapshot of your data. For example, if you or say a hacker broke into your computer system and deleted important files, there is no RAID level that is going to help you recover that data; but you can get a copy of the deleted file from last night's backups.

Since you say data availability isn't an issue, I think RAID1/4/5/6/50/60 are not necessary in your situation. Ask yourself this question:

"if your hard drive fails, 1) how long will it take to recover from backup, and 2) will the backup be fresh enough, which depends on your frequency of backups?"

If you plan to backup frequently enough (say every X hours), and it takes Y minutes to get your data from backup, is a copy of the data from X hours ago restored in Y minutes sufficient? If so, then you don't need RAID.

If you do need RAID, I would also take into consideration the use of RAID1. It's a much simpler configuration with little performance impact on writes, and possibly a performance improvement on reads (depending on the implementation).

RAID5 is okay, but you'll have performance degradation with only 3 HDD, especially if the RAID processor isn't powerful.

The less number of hard drives, the less your chances are of encountering a failed drive. Something to think about... in a data center with technicians/operators 24/7 and spare drives in a locked cage, having 10 drives fail every day out of 10000 hard drives isn't an issue. Having a drive fail at home, RMA'ing the failed drive, waiting for the replacement to come back can be an issue.

mdogg

I'm a bit confused by that part.

Right now I'm leaning towards no RAID simply because the critical data will be on S3

BLinux

"If you plan to backup frequently enough (say every X hours), and it takes Y minutes to get your data from backup, is a copy of the data from X hours ago restored in Y minutes sufficient? If so, then you don't need RAID."

maybe a couple of examples would help:

1. let's say I do a backup every 24 hours @ midnight every day. let's also say that if for some reason, my computer blows up, it takes me 15 minutes to download the data from my backup area. so, if this is for my personal emails, and my computer blows up at exactly 11:59PM, then my last backup will have lost any emails i may have received in the last 23 hours and 59 minutes. For my personal email, that might be okay... i don't get that much anyway, and if there was something important, i usually would know who to contact and say "hey, that email you sent yesterday, my computer crashed and I lost it, can you send it again?" So, in this case, do I really need the availability provided by RAID? No...

2. let's say I'm using the same backup scenario as example #1, but this time, I'm dealing with my online order logs for a small website I run selling SuperWidgets. if I lost 23 hours and 59 minutes of orders that I need to process, well that certainly isn't good and I wouldn't know who to contact to tell them "hey, can you tell me your order again? we lost our online order logs"... I'd probably have a lot of angry customers that paid for something I have no recollection because I lost my order logs. In this case, something like RAID1/4/5/6/50/60 would help me avoid this type of catastrophe in the event of a hard drive failure.

3. let's change up scenario #2 so that we're doing backups every 1 hour. but let's also take into consideration that it would take 3 hours to setup a new machine in the case that my computer blows up. so, my SuperWidgets are selling, but I'm really only taking in 1 order every 1-2 hours. So, if I didn't have a redundant RAID protection, and I lost some orders because i had a drive failure, I *might* lose 1 order (about 50/50 chance)... well, I might just say I can deal with that.... if I do lose an order, I'll likely get a nasty email from the unhappy customer and I can deal with him/her individually and correct the problem. So, if this was the case, maybe I don't need RAID redundancy. the other problem I have to deal with is, if my drive failure results in an inoperable computer that needs to be fixed and setup again... that would cost me 3 hours of time; is that acceptable?

4. so, let's say you're not into selling SuperWidgets. let's say instead, you're an amateur musician hoping to make it big. you do your recordings on your computer and yesterday you finally got a recording that was just right for your demo CD. well, if your backup happened before that recording was done, and your drive fails, you've just lost some pretty valuable stuff that you worked pretty hard for. maybe you can just re-record, or maybe that would be a lot of hassle and you wished you had RAID redundancy?

Does that help illustrate the point I'm trying to make?

There's a trade off you have to consider... using RAID redundancy does protect you in a way, but having more hard drives also increases your chances of encountering failed drives. That's why, for most home use, I recommend just using RAID1 with large drives; it's simple, it's easy, and you're protected.

Now, if you said, you really need a LOT of storage, or you really need very high performing disk sub-system (and really *need* could be just that you *want* it just for the sake of having it), then I would say go ahead and play with RAID5/6/50/60 (or even RAID0), but I would also recommend that you get a lot more than just 3 HDD.... if you really need it, or really just want to play with a disk subsystem that can do 500MB/sec, get at least 6-8 HDD, or get a disk array with 15-20 HDD...

DuckieHo

Just to throw my thoughts in....

Make sure to consider archiving as well as backups. Archiving is long term and securer but is more difficult to restore due accessibility. Backups are more for quick restores.

__________________
To answer most of your questions: (1) a fridge cannot cool a PC (2) 64-bit OS for over 3.4GB (3) If a PCIe card fits, it should work (4) Resolution, not screen size (5) If you have a question, it is not news (6) Report, not respond to Spam (7) Single-Rail/Non-Modular PSUs are not always better than Multi-Rail/Modular

mdogg

BLinux, that was an amazing explanation, thanks

I am still leaning more towards no RAID or RAID 1
Ripping all my DVDs and CDs took a really long time and its something I reaaally dont want to do again. Also the cost involved with backing up my music and movies on the cloud is just outrageous.

If I do RAID I want to get a decent RAID controller because I don't want to have to deal with 24 hour build times, and random failures. Or am I just worried about nothing?

Thanks a lot guys.

I'm really split....

__________________
Some stuff I made

Good Linux Users Don’t Talk About Microsoft

Really, I’m not out to destroy Microsoft. That will just be a completely unintentional side effect.– Linus Torvalds

BLinux

If you don't need fast access speeds to your data, consider getting a Drobo. This is a nice little consumer product that helps make things like RAID easier for the end user.

http://www.drobo.com/resources/drobodemo.php

I have no affiliation with that company, a friend has it and has said good things about it.