Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

If I were to use a flash storage device on a server, a portable one specifically, and backups written to it periodically every 5 hours or so, would there be much wear to the flash drive to the point of depending on it as a reliable fallback to data recovery? Since there wouldn't be a lot of writing involved all the time, the flash drive's lifespan isn't reduced that significantly right?

Or maybe somehow in addition, get the filesystem driver to use fresh blocks for all allocations instead of old ones to maximize flash cell lifespan?

Reason I want to do this is to avoid using up bandwidth for network backups, and there aren't any other local servers to perform such backups without overall internet latency. And that flash storage devices are cheap and easy to obtain.

NOTE: I'm no hired IT, I have a low budget and run a home server if that's any specific. Also I'm looking to backup are web directories and SQLServer DBs.

share|improve this question

migrated from serverfault.com Apr 30 '12 at 8:49

This question came from our site for professional system and network administrators.

2  
Hey what's with the -1? A don't see a similar question around, so I don't see why this question wouldn't be useful. –  chaz Apr 29 '12 at 9:35
    
I've no idea about the -1 but you have left out important information about how much data you're backing up, how often, etc. –  Iain Apr 29 '12 at 12:09
    
Oh yeah I totally forgot! Gonna update. –  chaz Apr 29 '12 at 12:12
2  
USB flash drives are unreliable. I work in a college and we see a large number of students who keep the only copy of their work on one that lose that work no matter how careful they are. As such, there's no way I'd trust USB drives even for home backup; if something is worth backing up at all, its worth backing up onto something reliable. USB flash drives should be used to transport copies of data and not much more in my opinion. –  RobM Apr 29 '12 at 12:54
    
@chaz Server Fault is for IT Professionals; Super User is a better fit for this kind of question. –  gravyface Apr 29 '12 at 16:22

4 Answers 4

I probably wouldn't rely on flash drives for server backups. I just wouldn't trust the reliability.

If you have no network storage then you could purchase USB hard drives for the backup. We use external USB hard drives for backup drives. We copy to a machine that has a large internal drive and then use a robocopy script to replicate to external USB drives that are swapped out each day and rotated. I have also done this for clients with drives directly attached to server.

share|improve this answer
1  
Well I mean I don't have a budget of +70$. I'm pretty cheap here. –  chaz Apr 29 '12 at 12:09
1  
@chaz: then you're not a professional sysadmin and you shouldn't be asking your question here, please read the faq. –  Iain Apr 29 '12 at 12:11
    
Okay, what if I imposed the question as generic? At least it'd be helpful for any similar questions. This question could be valid for a temporary backup solution, but I wouldn't honestly rely on this as a long term solution myself. –  chaz Apr 29 '12 at 12:14

If your client cannot afford more than $70 to backup data, look for a better paying gig.

That being said, the other answer is valid, they are unreliable and too small. Why not take a look at software like CrashPlan (free) and do versioned synchronization from one machine to another that that has a large disk just for backups from other machines. This will only help you with data unfortunately, not system state operating system recovery.

share|improve this answer

I've had some experience with failed flash drivers, and usually the main problem was the controller, not the flash chip itself, but I still wouldn't use it for main backup.

Flash chips usually have limited write cycles, so you can overwrite each 'sector' only times (~10.000). If you write one backup per day, that's a long, long time. If you write logs there, or copy something every couple of seconds (or have a /tmp dir there), that's not going to last long. Most newer controllers have wear leveling, and change the location of blocks (they have a couple of them extra), so the wear is distributed across the whole flash chip.

Using it as an extra backup, with sequential writes (no often overwriting), might be a good idea, but i woul definitely keep my "main" backups somewhere safe (hdd raid, or tape, or both).

share|improve this answer

What's interesting with this question is the shoestring budget. This makes it less off a professional server question. However there are some simple tests that can be applied to this kind of question regardless of scale of economy. (This is after all a site for giving / getting answers not comparing the size of each other's paycheck).

To directly answer your question. No SD will most likely be fine. Oddly a backup that completely overwrites the card is one of the least intensive uses of one. Its not the number of bytes you write, but the number of times you change an individual byte) What really kills them is letting a program use them for temp space (as some programs will automatically do the moment you edit a file directly on the flash drive). Remember that backing up every 5 hours is only changing each sector every 5 hours. It will wear, but after a long time.

Is this advisable? That's a simple question to answer: What is your data worth to you / your company / your client?

Think of a backup like an insurance policy. On the assumption you're going to replace this backup solution completely in 2 years. $70 is cheap for protecting the average home business data for two years. The price of a sizable SD card these days is almost as cheap as many of us pay for a coffee. So only consider using an SD card if loosing that server's data would be an annoyance (couple of weeks work wasted) and nothing more.

Make sure your backup solution is verifying each backup (SD wear should be detectable if you also store check-sums of the files) and swap between two or more so that if one fails you're not more a day or two behind.

Finally consider using a good quality flash card. Remember that photographers fill up and clear down these daily. Your suggestion of doing so once in 5 hours isn't that far removed from what the good quality ones are designed for.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.