Disk wear and tear 2

[xwb]

I don't know if I'm in the correct forum for this.

We have an application that runs on a server. This is the only application on the server. One thread writes files to a directory. Another thread copies them to another machine and deletes them. From what I can see, the files are always written to the same area of the disk. Sooner or later, that area of the disk will be unusable while the rest of the disk will be OK.

I was just wondering whether we should not delete any files until the free disk area gets down to 20%. That way more of the disk will get used and the disk might last longer.

Does anyone have any opinions on this?

 

[stduc]

There is no physical contact between the read/write head & the platters. Therefore there is no physical wear & tear. The heads will be moving to the directory tracks from time to time so the head stepper mechanism will not get stuck either.

In short - the disk will die when it dies. Using it more or less won't make any noticeable difference. Frequent power cycling will shorten a drives life as will over heating, excessive vibration, high altitude and a host of other environmental factors. But unless you are running that drive hot and and 100% throughput 100% of the time - your reported usage is just fine.

If your worried about the drive going down, maybe you should think of making it part of a RAID array. Either RAID5 or RAID1.

[navy]When I married "Miss Right" I didn't realise her first name was 'always'. LOL[/navy]

 

[wahnula]

Another thread copies them to another machine and deletes them.

Just curious, what would be the point of copying something only to delete it? Is it as backup that deletes old files then copies new ones?

Tony

"If it can't take it, I don't want it

 

[xwb]

After copying them to the other machine, it deletes its local copy. I suppose it is like a man with one leg called Fred. I didn't put enough words to give it an unambiguous meaning.

 

[edfair]

Holding onto files will result in more head movement and some additional heat. Stick with the present method.

Ed Fair
Give the wrong symptoms, get the wrong solutions.

 

[cdogg]

Like the others said, the wear on a drive will not increase in one particular area on the disk. It is the amount of "head movement" gliding over the platters that wears down the lubricants on the inside of the drive. At some point, they break down causing the head to glide above the platters at the wrong distance leading to bad sectors. In fact, drives can actually fail altogether without warning when the head actually makes contact with a platter.

So in your case, I wouldn't worry about it unless you notice the process taking longer and longer to complete over time. That would be a sign that the amount of head movement needed to complete the same task is also increasing. You might want to consider using a RAID array with fault tolerance like stduc suggested as a precaution if you're not already...

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[stduc]

There is no contact between the heads and the platters of a hard drive whilst it is operating. The heads "fly" on an air cushion about 10nm above the surface. The effect and aerodynamics that permit this to work were discovered by an Swiss scientist called Bernoulli. So lubricants don't come into it. The platter is covered with a monolayer of special lubricant, but this is to facilitate head lift off when the drive is powered up. This is why drives that are left running 24/7 tend to run longer (in terms of hours used) than drives that are turned off on a regular basis. When a drive is powered off the heads are moved to a special data free area known as the parking zone. They don't just land anywhere.
You can read more about it by clicking here.

[navy]When I married "Miss Right" I didn't realise her first name was 'always'. LOL[/navy]

 

[LawnBoy]

From what I can see, the files are always written to the same area of the disk.

I doubt that. The last OS I've seen that wrote the same file to the same sector on a disk was RSX running on a PDP-11. Modern OSes intentionally move files around on a platter to help with deleted file recovery. Even DOS 3.1 does.


"We must fall back upon the old axiom that when all other contingencies fail, whatever remains, however improbable, must be the truth." - Sherlock Holmes

 

[cdogg]

So lubricants don't come into it.

stduc,
But you said it yourself, LOL. They "help to facilitate head lift-off" and as your googled link continues on to say, landings too. Realize that the entire platter is coated with the lubricant. Air flow inside the drive relies on its presence across the entire surface of each platter. They don't just coat the area where the heads land and take-off from. Also, don't forget about the spindle and head motors. They have lubricants too.

I really don't think it's necessary to break it down to such a level of semantics. The point I was making before was that mechanical parts inside wear down. Slice and dice it any way you like. In fact, there are dozens of things that can go wrong both mechanically and logistically to cause failure. This link lists many of them (note the "Friction of Internal Parts" that I touched on earlier):

http://www.datarecovery.org/hard-drive-crash.html

Here, we see a company that specializes in Data Recovery say that "Stiction" is a common form of drive failure, which is defined here:

http://www.idsemergencymanagement.c..._data_recovery_labs_inc_/12_0/g_supplier.html

the read-write head assembly gets ``stuck`` on the disk media due to deterioration of the lubricant or because it has failed to retract to its rest (parked) position.

Think I'll rest my case on that note!!


~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[cdogg]

xwb,
Sorry to go off on a tangeant. I figure you've probably already found what you're looking for...

 

[xwb]

No problem - I'd forgotten that the heads don't actually touch the surface: I last did that 33 years ago in Uni. Nice thing nowadays: you don't have to be a weight lifter to carry a disk and you don't get a backache putting the disk in the drive. They are now so secure that there is no chance of a disk flying out and denting filing cabinets. Disks were dangerous things 30 years ago.

As for files being written around the platter, I'm not so sure. I've looked at the drive map after 1 write, and every day after about 500 writes and deletes and the map still looks the same. The blues are always in the same place. They never move.

Anyway, nice discussion.

 

[LawnBoy]

xwb,
You've got me curious, what are you using to look at the drive map? Is it mapping the actual sectors on the disk?


"We must fall back upon the old axiom that when all other contingencies fail, whatever remains, however improbable, must be the truth." - Sherlock Holmes

 

[xwb]

Diskeeper. I don't know if it maps the sectors or not. Just looks like it does.

 

[LawnBoy]

I've not used it, but a quick perusal of their website tells me it's a defragmenter. This is a higher-level file system function that is not looking at the physical platters on the disk.

I'm not sure what you could use to examine the physical disk these days. Way back when the dinosaurs still roamed the earth Norton Utilities had a raw disk editor, but I don't know of a modern equivalent.



"We must fall back upon the old axiom that when all other contingencies fail, whatever remains, however improbable, must be the truth." - Sherlock Holmes

 

[wahnula]

SpinRite?

Tony

"If it can't take it, I don't want it

 

[xwb]

There hasn't been anything around since IDE disks came out. It has been a long time since I've played with this stuff. The last time I played with an IDE disk, it told me it had 255 heads which was clearly a lie. I've got no idea what the disk configuration is at that level but I'm not really bothered anyway.

As for sectors, I think defrag tools all report the same sort of thing. As far as I can see, it is more or less a sector map. I don't think the OS moves sectors about unless it has some bad sector management. I've got no idea what windows does for bad sector management or whether it even has any.

If the OS writes to different sectors every time, then the positions of the used sectors will keep on moving. If it writes to the same sectors all the time, then the positions of the used sectors will remain the same. What I'm seeing is that the position is always the same which leads me to conclude that the system is always writing to the same area.

I'm just pursuing this out of interest. The system has been delivered and I can't do anything about it now but I may be able to do something on the next system which is planned for delivery in 2 years time.

 

[LawnBoy]

I think (but I'm no expert here) that what you're looking at is what LBA is reporting to the OS. LBA divorces the OS from the physical disk geometry.


"We must fall back upon the old axiom that when all other contingencies fail, whatever remains, however improbable, must be the truth." - Sherlock Holmes

 

[wahnula]

This is a fascinating educational thread for me, will someone please look at SpinRite

http://www.grc.com/sr/spinrite.htm

and see if it will do the requested job? It says it offers direct, low-level access to the hard disk.

Tony

"If it can't take it, I don't want it

 

[edfair]

Spinrite gave low level access up until the IDE drives locked us out. It was great for doing an "on the fly" low level format on ST225 type drives that had stepper motors for head movement and platters that were heat sensitive and tended to grow as they got hot.

If you would like to look at the disk sectors you could try winhex. And what you would do is look at the FAT and see how it changes as files are added and deleted.

I suspect that your operating system is adding the available sectors (clusters) back into the chain at the beginning. Otherwise the used structure would keep building away from the rim and you would start hearing the hard drive moving the heads to the data.

It has probably been 20 years since I last spent any meaningful time in a FAT and I can't remember how it worked then. But that was also when drives were 540mb or less and had smaller clusters and you could follow a chain in an hour or so.

Ed Fair
Give the wrong symptoms, get the wrong solutions.

 

[wahnula]

Spinrite gave low level access up until the IDE drives locked us out

Ed,

Could you take a moment to elaborate on this point, like what you mean by "locked us out"? I'm not the OP, just a sponge for knowledge...and archiving more threads as I get older. Thanks

Tony

"If it can't take it, I don't want it