A Backup Solution? | Hacker News Reader

A Backup Solution?

aaronsw.com

19 pointswglb24 comments

FWIW, Time Machine can actually be set up to do remote backups. You can even (p)re-configure the disk image if you want to limit the size, in case you don't want to partition your disk. I'll grant that it's a bit of a pain over the 3-click built-in solution, but it's what I'm using these days.

Enabling network backups: http://rajeev.name/blog/2007/11/03/time-machine-with-smb-and...

Limiting backup sizes: http://www.macosxhints.com/article.php?story=200711080201215...

I just bought a TimeMachine here since my whole house is now Mac and it finally seemed worth it. The initial set-up was a bit hairy: my wife's machine got caught in some sort of bad loop and every time we tried to restart her initial backup, it grew. First it was going to be 45 GB. When that failed and died, the next try was 90GB. We killed that, messed with a few things, tried again, and sure enough it was now going to be 135. Ugly.

Having said that, after wiping the TimeMachine and starting from scratch, everything has been perfect. And obviously all the backups after the initial one are very small and quite fast. So I suppose I would give it a B-.

http://www.tarsnap.com/

The blog post specifically mentions tarsnap in a list of applications that he then dismisses (they don't "just work").

You might disagree, but if you do, please say why.

I don't get his definition of "Just work" though. First he says that For Macs, there’s Time Machine, BackBlaze, Mozy, and Carbonite. But none of them seem to just work.

Time machine has got to be the easiest backup software I have used. That is the definition of just work in my book. Tarsnap looks cool but it is command line only so I can see why he dismissed it for a "just work" solution.

Edit: OK I went back and read his old article where he defines just work and Time machine does not "Just work" according to his definition.

I'd probably agree that it doesn't "just work" (in the sense that e.g. dropbox just works), but if you're reasonably comfortable with a command line and have used tar before, it's pretty straightforward.

The article does not support the author's 'just works' perspective. Its not much of an article at all but simply lists a few product he's tried on various platforms.

For me, I expect a certain amount of setup in any secure and fairly robust backup solution. tarsnap has very little setup for a person that loves the command line.

The problem with these sorts of services is that they require finicky clientside software.

The problem with real "storage" services (S3, Rackspace, etc.) is that they're ridiculously expensive--on the order of "buy a new backup drive every few months" expensive. Plus there's no way to ship drives to them or have them ship drives to you in an emergency.

I'd love to colocate a Backblaze pod in a datacenter and just rsync a few TB up to it. That way I can handle my own xattrs, my own bare metal, and not worry about finicky client software. Problem: $8k or so is way outside my backup budget...

Yeah, I know I could just colocate a $500 NAS, but it seems to wasteful to have so much CPU, power, etc., just for one backup.

If only we could round up a few hundred HNers to split a Backblaze pod...

WOW... that is impressive. 67TB in a 4U. They put a lot of thought into those servers.

http://blog.backblaze.com/2009/09/01/petabytes-on-a-budget-h...

Edit: Impressive service and great price point. It looks like the only thing that I would want to have as well is geographic redundancy too.

FYI, you can mail drives to S3: http://aws.amazon.com/importexport/

Device Handling $80.00 per storage device handled.

Data Loading Time $2.49 per data-loading-hour. Partial data-loading-hours are billed as full hours.

Yes, on the order of cost-of-drive.

For my servers (linux), I use tarsnap. For my workstations (OS X) I use SuperDuper.app to image my hard drive.

We've spent many hours evaluating different backup services -- we're specifically looking for one with a partner program that's appropriate for our business.

The good news is, most services now (claim to) offer encryption as part of their basic package.

For Mac users, Carbonite does seem to be the best service available. For Windows users, I'd recommend SpiderOak, who are cheaper in most cases than other services, and they seem to offer a reasonably good client program.

For potential resellers, in our opinion there just isn't a good one available yet. We hadn't found CrashPlan though, and so far it looks promising.

The problem IMHO is that the encryption is just that -- a "claim". You can't access your data on their servers, so you can't verify it. You can use your own password, but there's no way to verify it's being used. Backblaze's blog http://tr.im/BPfR for instance says that "IT" has access to my data without knowing my password.

One of the advantages (and disadvantages) to CrashPlan, for service providers, is that it allows you to run their server software on your own infrastructure.

But yes, I agree with you otherwise. I'd much rather trust cperciva with data, but unfortunately his system lacks some critical features for our uses.

I'd much rather trust cperciva with data, but unfortunately his system lacks some critical features for our uses.

If you haven't already done so, please send me an email listing the lacking features which you find critical. I can't guarantee that I'll add them, but it's always useful to know how many people need what features.

The good news is, most services now (claim to) offer encryption as part of their basic package.

Yet they still make n00b mistakes like "we recommend that you use the built-in encryption key (which everybody who downloads the client software has access to)". And when they don't announce their mistakes publicly, they still don't provide enough technical details (e.g., cipher block mode) to confirm that they're not making dumb mistakes behind the scenes. (It doesn't help when people refuse to make source code available, of course.)

http://www.haystacksoftware.com/arq/ would seem to fit his needs for all but the bare-metal restore.

But it's unclear whether he requires an online (offsite) solution or not. For local backup, Time Machine works fine for all but restoring hard links, fifos and devices, which may not matter in his case. You just have to remember to plug in your USB drive, or be near your Time Capsule pretty often.

I have been using http://www.jungledisk.com (acquired by Rackspace over a year back). You pay only for the storage you need (I prefer that over unlimited backup services), you can choose Amazon s3 or Rackspace cloudfiles for storage, client software available on all three platforms works fine and you can encrypt the data with your own passphrase.

http://www.hybir.com

full online backup

bare metal restore

windows only for now.

Interesting performance advantages can be seen in the screen casts here:

http://www.hybir.com/Action/HybirBackupAction.html

(blatent self promotion)

The main problem with Time Machine is how badly it "works" with File Vault... backing up the encrypted stripes only, and only when you aren't logged in to your Mac, which in my case is approx 0% of the time.

I wouldn't call that working badly - it seems to be about the only reasonable way to handle the situation. I just started working a logout at the end of the day into my schedule and never really thought about it since.

I'm curious what you would rather have it do… backup the content unencrypted, or force a logout every hour to do the backup?

Back in time + sshfs + cron job is working pretty well for me.

Try Tonido Backup. http://www.tonido.com/app_backup_home.html.

Works on Windows, Mac and Linux. It is free too.

Reply on news.ycombinator.com