Tape Drives vs Blu Ray

[pete376403]

Yeah, you're probably right.

About $10K for a 20 pack of Caviar SE Green Powers is the cheapest I can find (have to allow for some units in the RAID dying - we're hard on drives).

Plus the $10-12K for the SAN you're suggesting.

means you pay over $20K for the hardware.

And how much for a rock-solid SLA? Probably about $10K for a three-year contract.

$30K. I rest my case.

The Intel box without any drives retails for $5K5.
The SLA, well I don't know, it has a three year warranty and if anything fails, I change the parts myself (as in the case of one drive that carked) Hot plug etc. quite a nice unit. Our one is probably way over specced, dual quad-core Xeons, 8 gig ram, could have been cheaper.

[Gremlin]

The only thing that comes close is a hard drive in a removable caddy, but are you going to trust a courier to handle that (ie offsite storage) in a manner that will allow it to be re-used?

I wouldn't trust a courier with anything really... handle backups myself normally (for those that pay for it )

the issue with a hdd in a caddy is the speed... quite limiting compared to a normal sata/whatever interface, when you are dealing with 100's of gb.

riffer: as a general rule, best not to buy all your hdd in one pack, because if there is an issue with a batch... well... suffice to say I have heard (but thankfully not had the personal pleasure) of raids being lost when multiple drives failed at once. Space the buying out, to hopefully buy from different batches.

[scracha]

I wouldn't trust a courier with anything really... handle backups myself normally (for those that pay for it )

the issue with a hdd in a caddy is the speed... quite limiting compared to a normal sata/whatever interface, when you are dealing with 100's of gb.

Does anyone actually know anybody who's lost two drives in a RAID array at once or is this another of these IT urban myths?

Backing up the ReadyNAS, yep...USB takes ages and I've only got 3x320gb drives shoved in the thing. eSATA on a Synology aint too bad though....seemed about 3x quicker.

So in response to my original question. Tape is still useful for offsite backups using a courier that are over 50GB in size? How long does tape last and what are the limits on the number of times it can be read/written to. Is it standard practice to verify data on a tape drive after writing. Dunno much about tape at all so again, apologies.

[Gremlin]

Does anyone actually know anybody who's lost two drives in a RAID array at once or is this another of these IT urban myths?

I believe I do... a mate and I were talking about single drive failure a while ago, but iirc, he brought up multiple drives failing...

Its also dependent on what drives you spec, what kind of abuse you give them, and how they subsequently cope. Some applications are terrible on drives, and if the drive is vulnerable... it doesn't help.

[Ixion]

I've experienced a dead server when three disks failed in a RAID 5 array (two totally dead one flakey as).

But, I must say , it was at a client site , and a cooling fan had gone. I don't know how long elapsed between the first disk going out and the total collapse (not long, but could have been several days). We just got called when "The computer's stopped working "

[Sanx]

the issue with a hdd in a caddy is the speed... quite limiting compared to a normal sata/whatever interface, when you are dealing with 100's of gb.

Erm, this might be the case if you're dealing with an external USB / Firewire attached drive, but HDDs in caddies are the absolute norm for any decent-sized server; all hot-plug drives live in caddies of some description.

And as eSata enclosures are available for peanuts these days (I bought one for $45 the other week, and it had USB2 and Firewire on it as well), there's not even an excuse for not using one, unless you don't happen to have any spare SATA ports on a motherboard.

[pete376403]

The Intel box I referred to supports hot spare(s) among other things. Gives you one more chance to organise replacement disk, ie even after one disk died the hot spare took over, so it was still R5, business as usual. If another disk had failed in that time then it would have been in degraded model but still working. If a third disk had died then it would have stopped

[scracha]

Yeah, even the cheap(ish) NAS' units have hot swappable drives.

[Gremlin]

Erm, this might be the case if you're dealing with an external USB / Firewire attached drive, but HDDs in caddies are the absolute norm for any decent-sized server; all hot-plug drives live in caddies of some description

woops... true I guess... was sorta meaning the usb ones, not hot swappable/esata etc

technically... the ones for hot swappable are called drive bays or whatnot

[RantyDave]

Does anyone actually know anybody who's lost two drives in a RAID array at once or is this another of these IT urban myths?

I believe it. The drives were made by the same manufacturer, to the same design, were part of the same batch, were treated the same in transit, were powered up at the same time and had identical workloads until one of them failed. It's not surprising the second is not far behind.

I worked on a big (for the time) ISP rig once and we had two big (for the time) LDAP servers shared on a load balancer. One day the LDAP went down and there was much waving of arms and shouting about load balancers being a piece of shit. Upon picking through the wreckage it became apparent that one of the servers had lost the plot a number of days prior, the balancer had failed over to the other one and the thing that had failed was the script that told the admin when it was time to go press the white button. I think a similar thing happens with RAID's pretty frequently.

Dave

[Gremlin]

I think a similar thing happens with RAID's pretty frequently.

Depends on the raid, the features of the controller, what kind of capabilities re notifying there is (if its been set up ) etc. There is also the old fashioned, check on it, see if its popping anything up (if it can't email or something).

From experience, you generally have some weird issues pop up just before the drive finally dies (sometimes it may even struggle on for a couple of days), but the issues depend on how good the machine is with surviving drives dropping out. Some will crash, and you restart and they are fine... some, the users don't even notice a performance hit.

You sure as hell move fast when that drive dies tho... tis a lot easier sticking another in, and the raid manager going... ooh, is that my new drive, cheers, let me rebuild for you... as opposed to... sitting there rebuilding the raid from backups...

[Matt]

LTO 4 is 800GB native / 1.6TB compressed - or disk to disk backup is a good option - even "named" brands (e.g. IBM) would sell you a iSCSI/SAS/SAN array for ~$5k, with $700 ish for each 1TB SATA HDD.

RAID 6 (DP) would protect against 2 disk failure, but at the expense of useable capacity - worth looking at as recovery time is massive on 1TB drives!