Re: Looking for Backup solutions for Linux

[Ez-Aton]

Shachar Shemesh wrote: Ez-Aton wrote: 12 / 2 (=half a dozen) * 20GB (per volume) = 120GB You are right. I misread the original post. Don't forget that these 120GB will likely only take about 60GB (on average, YMMV, yada yada yada) of actual space on the backup medium, but I agree that it's a bigger monthly cost. I can't see any resonable priced internet line which supports such online backup. Depends on your definition of "reasonable". Mostly, it depends on "how much is it worth it to you to not have to manually take your data off-site". When we originally started to plan the backup service it was clear to me that people who want to backup 120GB of data are not my intended audience. The reality of things is that I have a lot of interest from precisely such clients. (no actual orders, but a lot of interest). For large/wealthy enough organizations, such TCP-based-method-of-moving-our-data-to-another-location, either in real-time, or daily, much like a backup (and almost anything in between) is a good method, and it solves almost every problem an organization can encounter. However, for the average place, in the smb category, in Israel, where connectivity prices for broader lines are proposterous, such an option is a nice-to-have-but-probably-too-expensive an option. They will always want to know, and then they will be so sorry they cannot afford the BW, and go for some other solution. USB IDE disks I do not recommend, for their low performance, and low reliablity. There is a reason why tapes are still the most common backup solution for a certain (and above) amounts of data. I agree that 120GB (as opposed to 20GB, as I thought before) suggest a tape solution. Still, the large up-front cost of the tape drive, coupled with the cost of each tape, make a hard disk solution seem appealing. Yes and no. Backup is all probablity. You play the game of chance, and you play it for your optimum amount of money. In a single-disk system, with no backup, there is a probablity of X that the disk might fail. There is much higher probability, Y, that some files will be deleted by accident. You add another disk, into a mirror, and you get X/1.5 that disk failure will kill your data. You add backup to the party, backing up once a week, and you make sure that you'll have a chance of X/10 that you will loose the whole data, Y/2 that some files will be erased beyond restoration, and you now add the Z factor of *how much data is lost*, which gets it all so more complicated. You backup once a day, you hardly change X, you decrease Y to be, maybe (all based on assumptions, for the matter) Y/5, and you change Z to be smaller (on a daily backup, I would expect Z/5, for the say). You add an off-site solution, and you decrease X, hardly any change in Y, and decrease Z, since you can rest assure that if you get to burn your office, you'll still have the data, to some extend. It can (and does) get more complicated, adding other letters into the pool, and it brings you, in the end, to the litte equation of less money, but lesser risk, or how much you'de pay to increase the survivability of your data. It's much like insurance, as you invest money to get better chance to gain something (your data) in case of an accident. After all this blah-blah, it's quite simple. It all depends on the size of investment the person who had the question post is to put into it. Using IDE disks, using custom kernel (if you're one of RH type systems fan) and relying on S.M.A.R.T to predict failure of disks (which happens, but it is rather rare. Usually SMART is as smart as any other prediction. And I'm sad to say I've seen so many SMART disks saying they're fine, with lots of bad sectors, head crushes, and more) has some appeal, as it is rather cheap, although risky (and we're here to decrease risks, right? That's what backup is for). If you backup during the day, you stress your system, so you would preffer to backup during night time, so you'll either have two such modules, or you backup every other night. Moreover, disks are not meant to be moved. They can be moved, but they experiance, even with head locks, etc, shorter life. Much shorter. And quoting you, Shachar - you wouldn't want the data to miss just when you need it. The alternate, more expensive, solution is a proven one. It is scalable - you get to see it in small and large orgs. You see it where industry hate to spend (you never need backup! You only need the ability to restore. Remember that. Your boss hates to pay for backup solutions, but he'll be all over you when data is missing, and you have no way to restore it). You see it where it has proven itself to be cost-effective enough solution to survive there. Alternate servers, disk containers, DVDs, mobile disks - none has been spread as much as backup tapes. Some other soltions are better for a specific custom environments, but most places just use tapes. It's the best cost-effective solution for minimizing the risk. Does your data worth 4000$ investment for three years time (assuming three years is a lifespan of a tape, which is a passimistic view)? 4000$, including SCSI controller, if you need, cartridges, replacing worn-out cartridges, and the tape itself. It is an estimation, based on experiance. Peace of mind is a key to better life :-) Ez Shachar Ez