Hers a storage solution with will blow your mind.
I built (with some vendor help) a storage/backup solutions that is as close to 100% redundant as you can get without mirroring the entire data center.

We have a 6tb EMC VNX disk array with dual power supplies and dual UPS backup power. No way it will ever go down due to no power....unless powers goes out and the 300gal diesel generator runs dry along with both  UPS's and both internal shut-down batteries all give up the ghost at the same time. Everything is hot swappable and redundant.

We back up 4 Raid5 configured servers to the EMC disks. Real time snapshot backups, nightly incremental, and weekly full backups. After the weekly full backup finishes the entire backup is cloned to DAT5 tape on another raid5 server which also backs itself up. On top of this each server also also has a DAT5 tape that we can spin data to whenever.

That being said in the past 10 years we've had about 5 or 6 out of about 30 hard disks fail in servers/storage. And they are running 24/7x365.

The whole damn thing is so complicated I don't even know if I could fix it anymore so it better never die. In fact as I'm typing this the backups are failing because we ran out of space and I have not figured out how to extend the back pool. All this hardware and it's failing because the software sucks and is not working like it should.

I've found drives, power supplies, monitors, etc tend to all go bad at about the same time (+ or - a few months).
When we install a batch of new PC's a few years later parts will start to fail in clusters. This tells me they are engineered  to fail. I mean, what are the odds every power supply dies in the same make/model/batch of machine with in months of each other?

thats the damned truth Eh?, my psu in main rig idk maybe 4yo, my spare is 3yo, had a drive work for 5 years plus when i sold quadzilla. still running along with 1tb drive in cali.

Quote from: john on June 05, 2013, 07:49:21 AM
Any backup is better than no backup.  I have a 2GB external drive that I push a backup every day to.

Oh yes - you have my vote sir.  The number of people that come whining to me to fix their troubled laptop drive that they have stored absolutely everything on for the last three years.. even with a MBR failure you can still typically recover something.  Laptop drives seem to cook and the controller fails, then you are stuffed.. 

Quote from: pliskin on June 05, 2013, 10:31:25 AM
I mean, what are the odds every power supply dies in the same make/model/batch of machine with in months of each other?

The odds are very good if you particuarly get a bad run of regulator capacitors especially.  All the components bought have a nominal MTBF - and due to the way they manufacture of components, you will get pretty much failure on the dot everytime. You buy MIL spec or aviation grade components, you don't get this problem - but you pay accordingly. Compare a voltage regulator for a bike made in China versus the Japanese equivalent - the Japanese one is twice the size, costs nearly twice as much, but lasts a lot longer.

You can usually spot the difference if you pick up a cheap "600W" PSU and pick up a high end 500W supply - the weight difference is obvious, the high end PSU will have better heatsinks, better capacitors and higher spec components.

PC power supplies (Buck converters) are very hard on the output stage capacitors in terms of noise - if there is inadequate high frequency filtering (or the filter caps fail due to heat), you will eventually cook the main regulator capacitors with harmonics from the switching.  You can buy more time before failure by getting a higher quality PSU, but PSU failure is not the issue to drives.

Mechanical drives suffer when the PSU cannot maintain the voltages levels required - effectively running the drive in a brown out state of power (draws more current and power dissipated goes up = cook the drive).

I guess I am used to seeing this in industrial applications everyday too - quality DC power is everything when you have devices that are operating to very close margins.  Some of the gear we work on is between 30 and 50 years old.  I will grant you that they are hideously expensive power supplies feeding this gear, but the payoff is very low downstream equipment failures = uptime.

BTW Pliskin - probably an overkill solution, but hey, awesome to get that much gear into a pile and working 

Holy crap what have I started my eyes have glazed haha. Goes to show how getting out of the game in 2003 has rendered me nearly illiterate today haha

Quote from: Kiwingenuity on June 05, 2013, 01:04:20 PM

Quote from: pliskin on June 05, 2013, 10:31:25 AM
I mean, what are the odds every power supply dies in the same make/model/batch of machine with in months of each other?

BTW Pliskin - probably an overkill solution, but hey, awesome to get that much gear into a pile and working 

I know, it is overkill. But I've been through a couple complete system crashes under unbelievable worst case scenarios where important data got lost. I do sleep a little better at night knowing if this one crashes I'll have good data....somewhere in there.

About 20 years ago, before this level of protection was available, I worked with a guy who did SQL database work. With a stroke of a key he wiped out over a million dollars worth of orders. Who's fault was that? His, or the hardware quys? Nobody knew. I don't want to be "that guy".

betcha that guy blamed hte hardware ppl. and hardware ppl blamed that guy. thats always fun to figure out and a great use for this particular smilie 

That would be a difficult call with a transactional database using SQL - so easy to drop a table in SQL if the commands were left available..

Thankfully we deal just with plain old data - LTO6 tape backups (incremental) with a full backup every week. stored offsite - This paid off when we lost our Christchurch office during the big quake.

Always good to be prepared 

Indeed sir. down the road as funds permit will be building quadzillas big brother using 2 1tb drives in a raid array

I think the biggest thing to consider in HDD versus SSD is this:

If a power surge cooks an HDD, 99% of the time it'll be just the onboard controller that's damaged. That means that the data can still be recovered by a professional. Essentially, as I talked about earlier, the data will remain on the platters indefinitely, even if the drive is badly damaged. Providing the platters themselves aren't damaged, you're golden.

The same thing, however, cannot be said for an SSD. A power surge would most likely destroy all of the data on the drive.

This is our backup scheme...

Readynas with redundant array at a datacenter as our primary fileserver.  A backup NAS, that pulls backups from the primary NAS.  A third external USB drive at the datacenter as a final emergency backup.

Servers in 2 locations help guarantee we can recover our data.

Quote from: Kijona on June 06, 2013, 08:38:02 AM
I think the biggest thing to consider in HDD versus SSD is this:

If a power surge cooks an HDD, 99% of the time it'll be just the onboard controller that's damaged. That means that the data can still be recovered by a professional. Essentially, as I talked about earlier, the data will remain on the platters indefinitely, even if the drive is badly damaged. Providing the platters themselves aren't damaged, you're golden.

The same thing, however, cannot be said for an SSD. A power surge would most likely destroy all of the data on the drive.

this is true. also figure lack of power over an extended time used to render SSD's empty or would chance it.

Quote from: yamahonkawazuki on June 06, 2013, 05:50:04 PM
this is true. also figure lack of power over an extended time used to render SSD's empty or would chance it.

Lack of power does - have seen a USB stick used for a product key fail after sitting three years in a safe.  SSDs use the same memory cells, so you run the risk of the same loss of memory happening.  Only have to plug it in to let it refresh though.

Have had to recover from platter once - that gets pretty expensive... at the time there was only a single provider and it cost several grand..

Quote from: Kiwingenuity on June 06, 2013, 05:55:21 PM

Quote from: yamahonkawazuki on June 06, 2013, 05:50:04 PM
this is true. also figure lack of power over an extended time used to render SSD's empty or would chance it.

Lack of power does - have seen a USB stick used for a product key fail after sitting three years in a safe.  SSDs use the same memory cells, so you run the risk of the same loss of memory happening.  Only have to plug it in to let it refresh though.

Have had to recover from platter once - that gets pretty expensive... at the time there was only a single provider and it cost several grand..

weve got many lappeis with ssd's ( mistly minis)  anyhoo in a pawn shop stuff is there for months usually. so every so often, ( once every 2 weeks)  ill plug them in for an hour or two.

Well in the end I now have a 2TB seagate external drive serving as my backup drive/youtube movie dump/ whatever else i feel like dumping drive with a 250gb WD drive as a backup backup for the seagate that's holding my precious family photos. Now I need a program that will when I want it to ghost a copy of my computer HDD onto the big seagate so I can get some sort of a backup regime going any suggestions?

crap 1tb main drive (size for speed)
crap 500gb externals (x2)
crap 160gb external with complete, full updated system and userspace clones
crap 40gb external with complete, full, updated system, no userspace clones
crap 16gb keychains with userspace prefs and passwords (encrypted)
siiiiiigh

all my photos and videos backup (ha!) automagically to mega, dropbox, picasa, and google drive

all my personal stuff get tard, gzipd, encrypted then backups (ha!) to mega, dropbox, and google drive.

occasionally ill look at the dvd burner and laugh. what a uselessly slow and complicated,  delicate pos. dvds might last 2 years, and have tiny capacity = worthless. that is, provided they don't get a gnats sack hair on them. or someone sneezes. or the manufacturer took a dump on it.

rsync, tar, gzip are my friends. I haven't lost data since the last oopsie with "rm -rf".

having a complete system clone, rsyncd weekly, with userspace stuff is priceless. 20 minutes of dumping one image back to the main box, and its like gnu again.


but I won't use anything but mac os and hackintoshes anymore. my understanding (past results) the next version (after 8)of windows might not suck, but we'll see.