New hard drive - best break-in procedure?

[elconejito]

It's never occurred to me to "burn-in" a drive, although my standard practice is to do full format, then write zeros. I've lately started doing checks with spinrite as well if i have the time. I also keep the old drive lying around for a little while before wiping it and reusing elsewhere. I'm also *really* paranoid about backups, so I make sure there is at least one, preferably two, copies of the data elsewhere.

 

[Blain]

Originally posted by: elconejito
I'm also *really* paranoid about backups, so I make sure there is at least one, preferably two, copies of the data elsewhere.

Do you use A Drive, SkyDrive, Mozy or some other online storage site?

 

[elconejito]

i use jungledisk for online backup. you pay a one-time $20 for the software, then you pay Amazon S3 rates for transfer and storage. I think it's about $0.15/GB, IIRC. I liked it because it uses Amazon's servers for storage, and they aren't going anywhere anytime soon. It is a slight bother to sign up since you need to sign up for an S3 account, plus a jungledisk acct. So for the less technically inclined I also suggest Carbonite which I haven't used, but have heard good things, and it's only $50/year for unlimited storage.

I haven't used skydrive, i thought I heard they give you 25GB for free, is that right?

 

[Blain]

Originally posted by: elconejito
I haven't used skydrive, i thought I heard they give you 25GB for free, is that right?

Follow my links...
Yes SkyDrive gives you 25GB free, but A Drive doubles that.

 

[pjkenned]

Originally posted by: alkalinetaupehat

Originally posted by: pjkenned
For the easy-to-understand example... look at how many people with <1 year old have DOA items versus ones that fail in month 11 on NewEgg. You see tons of DOA mentions, quite a bit of "failed after 1 week", but 6-12 month failures are not reported too often. I understand that there are reasons for this, but it is a good example of the infancy death curve. There are much better experts on this than I but that's the general concept.

Those people (at least most of them anyways) probably didn't load their drives for a week straight, so I wonder if perhaps simply running the drive for a week would be sufficient. While it doesn't test the seek needle mechanism as much, it does put the entire drive under a long-term running condition where the HDD has to deal with built-up heat and applied voltage without a shutdown period.

I think the best idea would be to use a drive as a backup of your backup, but only access data from that drive for the duration of the "break-in" instead of wherever it normally is stored so as to apply a semi-normal load with non-mission-critical data so that there is a very low risk of something being lost, damaged, etc.

Thoughts?

So what I do, is usually, do a full format, then error check, then just have the disks back each other up for 3 days (I never buy single disks anymore), then full format/ error check, 3 days of backup, format/ error check. This isn't the best method by any means but it is sufficient. Also, I make sure to turn power saving features OFF when I can, remembering to do so for the on board or add-in raid controllers.

In the old days, disks didn't have great power savings features so just letting them run at 5400rpm or 7200rpm was ok. Now you have to watch out for them. Also, heat is actually a good thing during burn-in. It is a standard practice to let components run flat out at much higher than normal temperatures to see how fast they fail. There is an equation that I don't remember off the top of my head that says something like if you run hardware at ambient + 20C you effectively halve the life of the equipment. A few days a few degrees warmer isn't really going to change the lifespan of the hardware, yet will stress it a bit more meaning you effectively are shortening the stress period required to put a certain amount of stress on the drive. Then again every drive I have (non-notebook) is running in a hotswap chassis(built in fans) or at minimum has a 120mm fan blowing over it. People make a big deal about hard drive heat, but in reality, the heat output even under load is dwarfed by a GPU/CPU and etc.

Also, burn-in doesn't make your drive last longer. What it does do is prevent a lot of failures that happen early on from destroying data. For example, if you took a drive out of its packaging connected it, and moved data to it, then it dies 72 hours later.

 

[terentenet]

Geez, what's the deal with burning in the HDD? Just plug it in, use it and if it breaks it's under warranty at least for the next 2 years. If you're too worried about losing data, get a 2nd HDD and set up a RAID1 array. Why waste a week of your life stress testing a HDD???? I really don't get it. By the time you finish stress testing, I have mine filled with data. I made a decision to not keep CDs/DVDs any longer. With HDD's being so cheap, I just fill the drive, index it and put it in the shelf for storage.

PS. Doing a burn-in to uncover a defective drive... it might do more harm than good. What do you do when you purchase your new car? Pop it in 1st gear and run @ red line for the first 2 days to see if the engine or transmission fail? Chances are it won't (since the red line is far under the actual upper rev limit the engine can take), but the stress you put on it will show in the long run.

 

[pjkenned]

Originally posted by: terentenet
Geez, what's the deal with burning in the HDD? Just plug it in, use it and if it breaks it's under warranty at least for the next 2 years. If you're too worried about losing data, get a 2nd HDD and set up a RAID1 array. Why waste a week of your life stress testing a HDD???? I really don't get it. By the time you finish stress testing, I have mine filled with data. I made a decision to not keep CDs/DVDs any longer. With HDD's being so cheap, I just fill the drive, index it and put it in the shelf for storage.

Most major drive makers are on at least 3 years for consumer drives and my SAS drives are 5 years, but warranty is never the issue.

I don't actively wait to stress test. Also, my WHS has 2x 1TB drives in Raid 1 and 13x 1.5TB drives in raid 6 + hotspare that handles all of my backups. My main PC has 8x 15k rpm 2.5" SAS drives in raid 5, 3x 15k rpm 3.5" SAS drives in raid 5, and 1x 1.5TB drive (purely as local backup to the raid 5 arrays). I make ample backups and redundancy I have covered. For my main PC, the merits of 700+ MB/s sustained sequential reads using raid 5 are such that raid 1 only gives a fraction of the performance.

Once you have a LOT of data, spanning years, losing data is not an option. Also, you really don't want to be rebuilding raid 5/6 non-stop once you are using 1TB + drives, even though I use all hardware raid controllers, it still takes a long time.

If you don't burn-in however, you run the risk of filling a drive, putting it on the shelf, and not being able to access it again when you plug it back in. Infancy failures on electronic devices tend not to happen on the shelf (although I guess they can).

The difference is that with burn-ins I am reasonably happy throwing drives into the raid arrays. Downtime costs $$$ and I have space left in trayless hot swap cages so my "waste" of time is a matter of minutes over a week.

Also, an index and shelves are great, but it is pretty hard to have remote access to that data when it is all offline. Then again, that method is very common for people who live/work in one town, and download/rip movies and applications that don't need to be used a lot. For that purpose, storage these days is super cheap, but you are really using disk drives for the same purpose people use tape back-ups/ CD-RW's/ DVD-RW's/ and Blu-ray recordable media for. The difference is the optical media tends to produce coasters when burned for infancy failure, while hard drives can take data, then lose it when you plug it in again due to infancy failure.

Bottom line is, as your data and time become more valuable, burn-ins are a great way to minimize the headache of a drive going down with data on it.

 

[taltamir]

Originally posted by: pjkenned

Originally posted by: alkalinetaupehat

Originally posted by: pjkenned
For the easy-to-understand example... look at how many people with <1 year old have DOA items versus ones that fail in month 11 on NewEgg. You see tons of DOA mentions, quite a bit of "failed after 1 week", but 6-12 month failures are not reported too often. I understand that there are reasons for this, but it is a good example of the infancy death curve. There are much better experts on this than I but that's the general concept.

Those people (at least most of them anyways) probably didn't load their drives for a week straight, so I wonder if perhaps simply running the drive for a week would be sufficient. While it doesn't test the seek needle mechanism as much, it does put the entire drive under a long-term running condition where the HDD has to deal with built-up heat and applied voltage without a shutdown period.

I think the best idea would be to use a drive as a backup of your backup, but only access data from that drive for the duration of the "break-in" instead of wherever it normally is stored so as to apply a semi-normal load with non-mission-critical data so that there is a very low risk of something being lost, damaged, etc.

Thoughts?

So what I do, is usually, do a full format, then error check, then just have the disks back each other up for 3 days (I never buy single disks anymore), then full format/ error check, 3 days of backup, format/ error check. This isn't the best method by any means but it is sufficient. Also, I make sure to turn power saving features OFF when I can, remembering to do so for the on board or add-in raid controllers.

In the old days, disks didn't have great power savings features so just letting them run at 5400rpm or 7200rpm was ok. Now you have to watch out for them. Also, heat is actually a good thing during burn-in. It is a standard practice to let components run flat out at much higher than normal temperatures to see how fast they fail. There is an equation that I don't remember off the top of my head that says something like if you run hardware at ambient + 20C you effectively halve the life of the equipment. A few days a few degrees warmer isn't really going to change the lifespan of the hardware, yet will stress it a bit more meaning you effectively are shortening the stress period required to put a certain amount of stress on the drive. Then again every drive I have (non-notebook) is running in a hotswap chassis(built in fans) or at minimum has a 120mm fan blowing over it. People make a big deal about hard drive heat, but in reality, the heat output even under load is dwarfed by a GPU/CPU and etc.

Also, burn-in doesn't make your drive last longer. What it does do is prevent a lot of failures that happen early on from destroying data. For example, if you took a drive out of its packaging connected it, and moved data to it, then it dies 72 hours later.

That happened to me before .

Specifically... I bought a new giant drive, MOVED the data to it from a few drives, then tried to make a raid1 from it... well, that was the plan at least, it failed halfway through with dataloss, I never do move operations anymore, only copy and keep the old drives for a while. (and everything is on redundant disks...)