big drives

Exactly. You need to look at *why* you are using the "storage medium" to determine the best way to implement it.

If the disk is being used in lieu of RAM (e.g., backing store for VMM) it will *tend* to see much different access patterns and rates than if it is being used as an archive/repository, etc. Backing store sees higher access *rates* as well as more random accesses than, for example, pulling files off an archive (where you can anticipate the next access based on the *previous* one -- esp if any care was taken in laying down the data tracks in the first place!)

Hence my use of higher performance drives (and smaller spindles) for "secondary" storage -- relying on gobs of RAM for primary -- and "commodity drives" for true "bulk" storage.

We back up to USB terabyte hard drives and use them as write-once media.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

And what ongoing *policy* do you have to verify their contents are still accessible (i.e., that the disk hasn't developed a flaw) *and* intact? Do you checksum the entire drive? All of its contents? Hope keeping TWO copies of that drive gives you *one* that is still intact, over time?

People have been making backups for decades. And, being *disappointed* by their backups, eventually, because they didn't invest in the entire solution, just a portion of it.

Massive redundancy. If 90% of the off-site drives die, we'll still be OK.

So far, the ones that I've checked were all OK.

Keeping dozens of monthly backup drives around the state is pretty good insurance. We have a lot of rolling daily on-site backups, too.

I think we should do the frequent backups to Dropbox, in addition to or instead of on-site, and still make the USB hard drive backups, once a month, to scatter about California.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Why don't you use SSD for that purpose? If you really need performance, use a PCIe SSD drive.

Worse is when you grab an old version, thinking that you got the latest.

For 3.5" drives, that might well be true. A 15K rpm 3.5" drive is mechanically much more demanding than a 2.5" drive. And since you can get much faster and much more reliable drives for less money, and you can also get brackets to fit 2.5" disks into 3.5" frames, it is extremely unlikely that anyone is going to make new 15K 3.5" drives.

Hard disks are not outdated technology, nor are 3.5" disks. But super-fast hard disks /are/ outdated - there is no point in them. SSDs are smaller, faster, cheaper, lower power, and more reliable - they win on every count. Your entire multi-disk raid could be replaced by a couple of big SSDs in RAID1 (or perhaps Linux RAID10) setup for a faster system taking less space and using a fraction of the power and noise. (Again, I fully understand that replacement is not necessarily a cheap or easy process - I am giving you the manufacturers' viewpoint here.)

Hard disks make sense when you want cheap GB's. SSDs make sense when you want speed. So new hard disks these days are typically 5400 or maybe 7200 rpm - 10K and 15K spinning rust is a thing of the past.

Technical glitches such as the Seagate plague make me hesitant to bet my farms on SSD. Other people can bet *their* farms first and we'll see what happens. At this present point in time it's best for me to "lead from behind" just like a science fictional hindmost puppeteer. :0)

Thank you,

--
Don Kuenz KB7RPU

My first one had 20MB but it cost about 10x of today's terabyte drives.

True but that all comes to an abrupt end when you are in the boonies, someone asks you to proof a presentation, the file size is 15MB and after finishing the first beer the download progress bar has managed to get to 5% when ... beep ... signal loss.

And then the rightful owner doesn't remember the password.

--
Regards, Joerg 

http://www.analogconsultants.com/

There is always a learning curve and SSD isn't the right technology for long-term storage (yet) but for a fast drive for computations, what's the problem? If it fails, so what?

They're also great for laptops, for obvious reasons.

SSD deployments into non-mission critical applications are actually quite old school to me at this juncture. Regardless it's still best for me to "lead from behind" and let other people bet their own mission critical farms first. :0)

Thank you,

--
Don Kuenz KB7RPU

Don't you mean a floppy drive?

--
Never piss off an Engineer! 

They don't get mad. 

They don't get even. 

They go for over unity! ;-)

No, I mean a fixed-head 4U rackmount drive, for a PDP-11. It cost about as much as two Chevrolets.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

What's "mission critical" about a cycle server? Add another.

A good file search engine is a joy to have. :-)

A good file search engine is a joy to have. :-)

My server farms, plural, each contain multiple servers. There's at most one cycle server in the mix. That particular cycle server is not mission critical.

The remaining servers of interest host mission critical applications for unaffiliated enterprises. It's just like the name says, when a mission critical server fails the enterprise fails. That's why it's best for me to "lead from behind" and let other people bet their own mission critical farms first. :0)

Thank you,

--
Don Kuenz KB7RPU

OK, but you haven't addressed the point. What's the purpose of a 15K drive - at all. For a cycle-server, reliability isn't critical so SSD is a natural solution. The disk in a byte server needs more reliability (RAID helps a lot) but not so much speed. A slower, high-density drive will have the same, or even higher data rate as a

15K and an SSD has *far* faster access. What's the point in the 15K? My bet is that's why they haven't improved much.

One of my vendors supports thousands of dermatology and gastroenterology practices across the nation. They insist on 15K. They won't even give me the time of day without it.

"You can't argue with success." And that's why 15K works for me too. Remember my "lead from behind" policy. :0)

Thank you,

--
Don Kuenz KB7RPU