Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

news.ycombinator.com

Post reply on HN
Print view View original

11–18 of 18 posts

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#11

Post by gcv »

jasonkester wrote: post #8

S3 does this for a living. I don't. S3's entire business revolves around never losing anybody's stuff. They have hordes of smart people working on the problem, and they have an architecture that makes it really hard to lose anything by accident. My business, on the other hand, revolves around letting people draw cartoon testicles onto other people's powerpoint presentations in the pretense of a "web meeting". Which o…

S3 does this for a living. I don't.

So does Carbonite, and, as you can see in today's news, it also lost people's data. I strongly suspect Amazon does a better job than Carbonite with both software and infrastructure, but still. Mistakes happen, bugs happen, and data gets lost even by smart people working for solid and reputable companies.

S3 only backups = eggs in one basket. It's a terrific, strong basket made of titanium and suspended on aircraft cable, but it's still just one basket.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#12

Post by iamelgringo »

My site, cuuute.com is hosted on EC2. I use Elastic block for database storage, and I back both that and my EC2 instance to S3 on a regular basis.

I've been using that as my hosting for 2-3 months, and I couldn't be happier.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#13

Post by lyime »

charlesju wrote: post #5

I trust S3 and Mosso Cloudfiles more-so than my own single-failure HDD. They have a lot of redundancy built into their system, and although every system has risks, its risk are far less than our person undistributed implementations of file storage.

It pretty much comes down to this. No one can guarantee 100% redundancy, not even Amazon. Although their setup and infrastructure will probably reduce overall MTTF (mean tike to failure) in comparison to a small hosting company or your personal backup solution. Very simple math can show that their system is probably more reliable then others, yours or mine.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#14

Post by sgoraya »

gcv wrote: post #11
jasonkester wrote: post #8

S3 does this for a living. I don't. S3's entire business revolves around never losing anybody's stuff. They have hordes of smart people working on the problem, and they have an architecture that makes it really hard to lose anything by accident. My business, on the other hand, revolves around letting people draw cartoon testicles onto other people's powerpoint presentations in the pretense of a "web meeting". Which o…

S3 does this for a living. I don't. So does Carbonite, and, as you can see in today's news, it also lost people's data. I strongly suspect Amazon does a better job than Carbonite with both software and infrastructure, but still. Mistakes happen, bugs happen, and data gets lost even by smart people working for solid and reputable companies. S3 only backups = eggs in one basket. It's a terrific, strong basket made of t…

So does Carbonite...

Good example - when I read the story I was not surprised. It was bound to happen to at least a company or two. I used carbonite for a bit before I went back to my own server (w/nightly backup). I thought I let my paranoid tendancies take over, but I just never felt good backing up my data to carbonite, et. al.

I also agree that S3 is much more failproof but would still rather keep things on my own server. I may be doing different in the future though, but for now there is no place like my own server!

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#15

Post by JungleDave »

bbuffone wrote: post #10

S3, Mosso and other cloud providers are not an appropriate back up mechanism. These are good for sharing files and a temporary storage system. The only legitimate backup is a physically stored disk. S3 doesn't provide versioning nor deletion protection. The problem is I hear about these new "Cloud" storage companies claiming backup, but when asked what do they do. They rely on Amazon to move the files into different…

Yes, that's a good example of why not to use S3 as a simple backup destination. However software like Jungle Disk on running on top of S3 adds versioning and deleted file retention to make it act like a real backup system.

Having a local backup is great too, but that won't protect you from fire, flood, or theft in many cases.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#16

Post by bbuffone »

JungleDave wrote: post #15
bbuffone wrote: post #10

S3, Mosso and other cloud providers are not an appropriate back up mechanism. These are good for sharing files and a temporary storage system. The only legitimate backup is a physically stored disk. S3 doesn't provide versioning nor deletion protection. The problem is I hear about these new "Cloud" storage companies claiming backup, but when asked what do they do. They rely on Amazon to move the files into different…

Yes, that's a good example of why not to use S3 as a simple backup destination. However software like Jungle Disk on running on top of S3 adds versioning and deleted file retention to make it act like a real backup system. Having a local backup is great too, but that won't protect you from fire, flood, or theft in many cases.

I wasn't talking about zipping up your files on a CD and putting them in a draw. There are companies like IronMountain, which give you off-site physical storage.

I am sure the deletion protection Jungle Disk provides is good but what if somebody internally deletes a S3 file or directory of files they are gone for ever. This isn't an S3 problem just a issue with the strategy.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#17

Post by JungleDave »

bbuffone wrote: post #16
JungleDave wrote: post #15

Earlier quoted context omitted.

Yes, that's a good example of why not to use S3 as a simple backup destination. However software like Jungle Disk on running on top of S3 adds versioning and deleted file retention to make it act like a real backup system. Having a local backup is great too, but that won't protect you from fire, flood, or theft in many cases.

I wasn't talking about zipping up your files on a CD and putting them in a draw. There are companies like IronMountain, which give you off-site physical storage. I am sure the deletion protection Jungle Disk provides is good but what if somebody internally deletes a S3 file or directory of files they are gone for ever. This isn't an S3 problem just a issue with the strategy.

Someone at Iron Mountain or Fedex could lose your backups too (accidentally or maliciously). Neither is too likely, nor is someone internally deleting data at S3. That said, if you have data you can't ever take a chance to lose, don't store it in only one place ever - not just on S3, not just on a USB drive, not just in off-site physical storage. Keep at least two copies, maybe more. One of the things we're working on is allowing you to backup to multiple cloud providers - so even in the unlikely event of a catastrophic cloud failure you'd have other copies available.

Re: Ask HN: Do you trust Amazon S3 or Mosso Cloudfiles not to lose or corrupt your data?

#18

Post by rs »

They are as safe as any other service provider. Ultimately, its always good practice to:

1. Do your own backups

2. Routinely test that you can recover from these backups

I would argue that point (2) is much more important than point (1). I do try to do that at least once a month to ensure that there aren't any bugs in the backups, including any missing parts of the infrastructure.

For mirroring really large data, rsync is a viable solution.

Edit: I do want to add that performing your own backups is really subjective and you might need to ask yourself - what's the cost to me/my business/my users in the event that I can't recover from backups and or my provider failed in their own reliability (for e.g. Carbonite)

Post reply on HN