Not too long ago, Amazon released their Simple Storage Service (or "S3" for short). It provides a hosted storage platform which developers can build all sorts of applications on top of. Smugmug, a popular photo sharing web site, is using it to store and host pictures.
I've been considering using S3 as the backend to an on-line backup, since I'd been beating that for a while (see: Swimming Pools and Hard Disks and Cheap On-Line Storage Coming Soon).
In a few days I'll write about how to do this--I'm only partially through the process right now. But right now I want to lay out the motivation for doing this.
The Cost of A Home Server
Amazon's pricing model is pretty compelling. The current rate is that I'd pay $0.15/GB monthly for the data I store. Data transfer costs $0.20/GB.
My home server is a 2.4GHz Pentium 4 that contains 3 250GB SATA disks in a RAID configuration, plus an 80GB boot and OS disk. I decided to measure its power use using my Kill A Watt and found that it consumes roughly 120 watts of power with the CPU idle and disks spinning. That's about 80 kilowatt hours every four weeks. According to PG&E's Residential Rates, the average cost of electric power in San Jose is $0.16247 per kilowatt hour.
My home server costs me a minimum of $13 every four weeks just to leave powered on and idling, or $170 per year. That doesn't count the roughly $700 I sunk into the disks and the other $700 I likely spent on the motherboard, CPU, case, RAM, and so on. Remember, these are absolute minimums, since the CPU does consume more power when it's actually doing work.
So if we assume that I spent $1,400 on the server and would keep it for 5 years, that's another $22 per month (4 weeks) of costs.
If you're keeping score, that's $22 + $13 = $35 every four weeks just to have backups of my stuff (most of which lives on servers in a few datacenters).
That puts the total cost for 5 years worth of backups around $2,275 assuming that no hardware breaks.
The Cost on S3
My backups require about 125GB of disk space today without compression. That'd cost me $18.75 per month to store on Amazon's S3. Let's further assume that I increase that by 1GB/month for the next five years (mostly photos) and transfer about 2GB every week doing backups (log files, mail, and other temporary stuff is much of that).
2GB every week is 8GB every four weeks, which costs another $1.60 every four week "month" for a total of $20.80 per year or $104 over 5 years.
Assuming that growth rate has me up to 190GB five years from now. Let's call it 200GB. If I'm growing at a constant rate, I can use the average of 200GB and 125GB, which is 162.5GB. Multiply that by 13 "weeks" and 5 years yields $1,584.
Adding it all up, if those guesses are right and we assume that Amazon's prices don't fall (they certainly could in a few years), I'd end up paying $1,688.
In other words, switching to S3 could save me $587 over five years!
Other Benefits
It's clear that going with S3 could save me money both from a reduced electric bill and not having to buy backup hardware (server and disks). But why else might I do this?
- Availability. It's less likely that Amazon's service will go down when compared to my home server and residential grade broadband service.
- Speed. If a remote server dies, I'd need to push all the bits there from my artificially slow DSL or Cable connection at home. Using S3 means I can restore faster.
- Simplicity. This is one less Unix box I have to spend any time administering. Even if it's only 5 minutes every week or two, that all adds up.
Again, I'll write up the process (tools and stuff) in a few days (or weeks).
Update: The economics are even better than I thought. It turns out that I mis-read some du output last night. I really only need half the space I thought I did. That is, ~75GB today instead of 125GB. Hmm...
Posted by jzawodn at October 03, 2006 08:51 PM
Great writeup, and I can't fault the math. While I haven't made the leap myself, I can certainly see the motivation.
And you left out one thing: the cost of S3 is likely to decrease over time, both in terms of cost of bandwidth and cost per unit of storage. Which means that your total 5 year cost could actually be significantly lower than your projection.
Worst-case scenario is that you migrate to another offsite host at a similar price point. The cost of *their* disks will always be going down, while when building it yourself you have to sink the entire cost up front.
But, and this is a big *but*, you did already sink that cost up front, didn't you? Not that you can't repurpose those disks, though your plan is even better for someone who hasn't already built their own backup solution.
PS, if your backup solution includes the wonderful rsnapshot utility, I'd love to hear it. I'm backing up remote servers to home via rsnapshot and if I could get it to work to S3, then I just might follow your lead.
...having your data when your net access is down: priceless
Ok, that was a bit facetious. I think I'd use it though as a supplementary to my onsite backup. Yahoo is not likely to go away any time soon, but then one might have thought the same of habeas corpus.
Brent: you missed the part about me having redundanat network connections at home (DSL and Cable). So it's extremely rare that I cannot reach the net.
DC has it right.
Your true cost is $845 for your at home solution, because the server is a sunk cost.
So S3 is actually almost twice as expensive over the next 5 years at $1688.
Still interested to see how you make it work. Don't let my fancy economics degree spoil your fun.
Actually, I simplified a few things. I'm pretty sure the server cost more than what I claimed, but it's about 2.5 or 3 years old now. So much of it is "paid for" in that sense.
However, my electricity is more expensive than the PG&E numbers because I currently go over the baseline they've established. There's a funky tiered pricing model at work that means that the price of electricity doesn't increase linearly. Bringing my consumption down quite bit has the greater than expected affect on my monthly costs.
Plus there's the issue of my personal time not being accounted for (I'd spent 8-10 hours messing with the hardware when I initially put the server together). And my house is much quieter with the thing powered off too! How shall we value that? :-)
However, I think my point is still valid. And, like DeWitt said, Amazon's price may very well go down over time.
I'm going down this path -- just laid some of the groundwork this week. I'm trying to tame the profusion of SLEDs (single large expensive disks) in firewire cases that keep two local copies of 50 gigs of important stuff and a hundred or so less important. I've figured out that I buy a drive a year for backups, and for roughly about that price at S3, I can have offsite backup and one fewer noisy thing with blue leds in my office.
What I'm not sure about is if some of it makes more sense to have backed up at someplace like dreamhost, which is a far cheaper (cause it's sunk cost) and good enough for most things.
I've considered options like this, but how do you know S3 will be there in 5 years? A9 is going/gone. Or, prices might change over time (up or down, who knows). Policies could change as well. Not to mention being unable to use standard tools like rsync, rsnapshot, etc.
Depending on the sensitivity of your data, it may be more cost effective to go in with friends on a machine in a hosted data center and backup to it.
I totally hear what you're saying about power costs here in the bay, though. The tiering is crazy. At least winter is almost here, when prices go down.
Funny you should post this tonight, I just ran some numbers involving S3: DreamHost vs Amazon S3
Ug, no HTML tags:
DreamHost vs Amazon S3:
http://joseph.randomnetworks.com/archives/2006/10/03/amazon-s3-vs-dreamhost/
Interesting idea, in addition to the critiques offered already I'd like to point out that your data storage needs may not increase linearly over time. They may suddenly increase exponentially -- perhaps if you start working with video.
I also see the upstream costs factored into your calculations, but not the downstream costs (for when disaster strikes). Of course, hopefully the cost of this is negligible!
Nonetheless I'm interested in how this turns out; I'm really interested in a total secure personal backup solution that scales to terabytes (*hand waving*) for my digital assets; all of my *really* important data like contacts, e-mails, and source code all already exist in many places, and probably barely top 2GB at most.
Hehe, I can really feel for the math in the blog post. When I am really interested in trying something out, I'll also calculate and fudge the numbers until I have it black on white, that it is clearly an economic decision and not my geek-drive that's making me spend time and money on that new thingie. ;)
So, if I really wanted to switch from S3 to an additional local machine, I would fudge the storage estiamtes up (like Mark Pilgrim calculated a few weeks ago, with lots of storage needs, S3 doesn't look so good anymore compared to a local solution), or I would calculate the energy needs through with an old/used laptop (burns only around 30-40 Watt), or take the initial investment costs of a relatively small and cheap box (you can actually get superp backup boxes for a lot less than 700$, and by your own calculation you only need 2 disks, not 4 for you local RAID, dropping the cost for disks to half). Then I would compare the S3 with the actual capacity I would have locally, not justthat which I wil likely need. And of course I'd slip in the bandwidth costs for S3. So I guess if you wanted to - you could make that local solution win the math as well... ;)
hi Jeremy --
Good timing on this post! I've been considering this option more and more, recently, too.
The biggest issue for me is the one Chip Turner raised -- Amazon's commitment to offering this service over the long term. Still, I suppose if it does go t*ts-up in a year or two, we will hopefully have enough warning to download the account's contents beforehand...
BTW, s3sync now sounds quite promising:
http://developer.amazonwebservices.com/connect/thread.jspa?messageID=44471
It says it doesn't do "incremental backups", but I hope that means it just can't deal with rsync-style partial changes inside files, instead working correctly with unchanged entire files. Haven't checked yet though.
How about the opportunity cost of the time spent reimplementing rsync? For me, that'd be pretty significant, which is why I'm still using my own host as a backup destination.
Another thing -- Amazon say in the FAQ that they'll have 99.99% reliability and availability. However, I haven't seen any contractual guarantees that if they somehow just lose your data, they'll do more than say "oops, sorry"...
At least with my own RAID, I'll have myself to blame. ;)
I have been thinking about this also, the biggest + for me would be to get a offsite backup by default to prevent things like fire destroying my computer and my backup.
Have though about buying another external disks and then bring one to work (and switch them every other day).
But knowing that this process would involve me as one failure points I dont think that it would work. :-)
Jeremy,
This all makes sense, but I'm curious, are you at all concerned about issues of privacy or security? While the photos may be a bit innocuous, backups often contain personal and financial data that I'm, at least, reticent to store online.
I'm with Eric on the Dreamhost option.
For $19.95/month with a $19.95 setup fee you get 200Gb of disk space (which increases by 2Gb every week) and 4Tb of transfer allowance (which increases by 32Gb every week), and the price never goes up.
Dreamhost is sounding like a better option than S3 for me for that reason, PLUS you get a shell, SSH access, unlimited MySQL databases, and all kinds of other stuff.
I'm not affiliated with them at all, just a very happy customer.
Preface: I am a current Dreamhost user.
Ok... that being said... search around on some of the problems Dreamhost is having right now. Slow service is one of them. For regular backups, that might not be that much of a problem. But it is another factor to consider.
Scoble was recently talking about the problems with Dreamhost as well. He has a few links to problems folks have had (check the comments).
I tried this a few months ago with the Mac client...
i def. think its the way to go... If Google can drop Points of Prescence storage datacenters in each region... they might be able to take Amazon's idea and make it actually work...
Im interested to know what your upload performance was like...
it was a dog trying it on my broadband...
Suresh
You neglect to mention one thing: The amount of time it takes to do network backups. It took me days usings S3 and there were a lot of errors...
http://www.informationweek.com/blog/main/archives/2006/09/the_achilles_he.html
Until bandwidth increases substantially, I just can't see remote storage being widely used by individuals with a large amount of data to backup.
Also, it needs to be a lot less prone to errors that make sense only to the command-line set.
We spend .07649 per kilowatt hour here Nashville. That worked out to $1755. This is $520 less than his amount of $2275 BUT still $67 about what you would pay Amazon. So you're still saving about $17.00 a year over 5 years based on today cost assumptions. Now, your other advantage is that the Amazon backup is accessible to you from anywhere in the world. Many home backups (Harddrive, DVD, CD, god forbid floppy disk) probably aren't.
But as always, saving money doesn't always add up when your placing your data and hence security in someone else's hands. I can here the NSA right around the corner. If you're saving family pictures and old records to backup, fine. But data sensitive proprietary information, no way.
Eric, I think the point wasn't specifically that DreamHost are cheaper, but that many hosting providers are.
Even if you don't like DreamHost, there are quite a few reliable shared hosting companies out there.
And S3 hasn't been up for long enough to know how they compare. They may be great and reliable, or they may have their own hiccups. It's not like Amazon is a data storage or hosting company with years of experience. Most of their experience lies in other areas.
And even much more expensive hosts still come out cheaper than S3 if you intend to have more than minimal requirements.
Heck, Get an account on both DH and two-three competitors, and a program to synch all three of them when you upload files (or at regular intervals). You get redundancy, off-site backups, and you can still come out cheaper than S3.
Add to that other services you'd receive in addition to just data storage, and S3 looks less than appealing.
Beyond the flexible pricing, which seems to be only attractive for low usage, what other advantage does S3 have over hosting providers?
Is their API that much more comfortable than FTP/WebDAV/whatever? If so, I expect it's only a matter of time until something similar will come along as a wrapper, and you can have the same comfort with regular hosts.
You might want to check out Carbonite
$5 a month for unlimited backup. You have to run there client and its only for one computer so its not quite a flexible but the price is right.
Sencer:
Sure, I can push the numbers around a bit to prove whatever I want. But I didn't have the outcome in mind when I sat down to gather these.
That doesn't mean they're *correct* but they feel close enough to me. :-)
Matt:
Sure. I'll know how that works out soon. :-)
I just finished recovering from a disk failure. For a couple of years I have backed up to my own server, racked in a co-location facility. I did not see the problem with my design until I actually needed to use it.
While my backup data was available to me, the latency getting it back into my house was excruciating (I have Verizon FIOS, with fat pipes in and out). I finally gave up, and drove to the co-lo.
I have since gone the opposite direction you have, and last week ordered a NAS to stick in my basement.
Steve
Stickan:
You hit the nail on the head. Backups *absolutely* must be automated. Otherwise they rely on my lazy human nature.
I've made that mistake before.
Thomas:
I'm not ignoring the issue at all. In fact, I wrote about it at length before and linked to that post from this article: Swimming Pools and Hard Disks
http://jeremy.zawodny.com/blog/archives/005820.html
But anyway, most of the data I'm backing up is *already* out of my house. It's on co-located servers in datacenters around the US.
I've also been thinking of this except more of an offline storage for photos and movies.
I've been looking into MediaMax services: http://www.mediamax.com/webservices/ComparisonChart.aspx.
They are quite a bit cheaper, especially for transfer costs.
-Dave
The biggest gain is high speed restores to any location.
If it's your home server it's still going to be running 24/7. So your power savings aren't going to be 100%.
If it was just a backup server, it doesn't need to be running 24/7. Wake-on-lan/scheduled startup from bios before the scheduled backups should result in more significant savings than from S3.
It doesn't make much sense out here in India though since my 1Mbps DSL line costs more than 3x the electricity bill for my whole apartment!
My home network uses around 0.3KWH, switching to a TFT display + a more energy efficient processor for the server would result in more significant savings.
I thought about doing something like this, then picked up a SimpleTech Simpleshare 160GB NAS thing, which I can plug external USB drives into. Total cost so far? $90. If I want to add disks later, I can do that for the price of a disk + enclosure. Even better, the thing will do RAID-1 and act as a print server.
Not that you have anything to hide, but remote services like Amazon are typically willing to give your data to government, law enforcement, defense lawyers,etc., with nothing more than a subpoena. (Which isn't that hard to get.)
Getting access to the data on your home PC would typically require a search warrant, which is a higher threshold.
There are two problems with S3, both in their licensing agreement. The first I haven't quite figured out, is in section 1.1, where you cannot exceed 1 call per second OR send files > 40K. I can't make out whether or not you can send files > 40K if you have less than 1 call per second?
Second problem is section 1.8, where Amazon can cancel your service at ANY time without ANY warning.
Neither of those clauses are quite what I'd be looking for in a backup solution.
The final problem for most people is the upload speed of their connections. Most are limited to 800Kbps, which translates to roughly 14 hours upload time for 5 GB, assuming you actually get 800Kbps.
I'm a big fan of amazon's new services. While I'm not sure I'd pay this much for my personal backup needs (i'm really bad about personal backups anyway) we have explored them quite thoroughly for some of our business applications.
More amazing than S3 is EC2, the elastic computing in the cloud. Slap up an instance, point it at your s3 data, and you're likely to replace your home servers all together.
I am with the others on DreamHost. With their recent boost in storage and transfer, it is a great deal. I am thinking about doing encrypted DMG images from my Mac and putting those on DH. Or creating a remote encrypted DMG image and using it over the internet via Webdav. For $8/mo that is not a bad deal. Plus, like the others said: ssh, mail, web, and mysql hosting is also included for the price.
Give JungleDisk (www.jungledisk.com) a whirl. I use it to back up all my photos to S3. About as simple as it gets *and* it encrypts :-)
I'll second jungledisk.com. I've been backing up all my personal data to S3 for over a month now. Many gigs later, my monthly statement was all of 63 cents. That I can live with.
Yeah, I've seen JungleDisk before but just gave it a closer look. Nice.
Dreamhost is even cheaper than Amazon S3, I think. Some folks are interested in hacking on Brad Fitzpatrick's "Brackup" --
http://search.cpan.org/~bradfitz/Brackup/
and making it use Dreamhost (well, any WebDAV or SSH-accessible storage, I'm thinking) as a storage node. The upside of Amazon S3 is you can buy more or less bandwidth or storage a la carte as your needs change, but Dreamhost is ridiculously cheap.
I am also a DreamHost customer. I am the one Scoble recently mentioned about having problems. I wrote all about my recent experience at http://nightmarehost.blogspot.com.
To summarize: Someone uploaded an AVI of Pirates of the Caribbean to the 'incoming' directory (which is configured to not allow downloading) on one of my anonymous FTP sites. There is no sort of notification for anonymous uploads, so I didn't know the file existed. But DreamHost didn't care and cancelled my account immediately without any warning and I had to fight with them to get it re-established after twice being told that my account would not be reinstated under any circumstances.
So I would not trust ANY of my private data with them. What if you uploaded something for your own personal use, such as a DRM-free copy of CD you owned? You could possibly wake up one morning to find yourself "permanently" locked out of your account, just as I did.
I'd like to see a solution that addresses what I refer to as "Propagation of Corrupted Data." Just wrote it up at http://www.blogarithms.com/index.php/archives/2006/10/04/personal-backup-on-amazon-s3/.
Well, having tried so many different esoteric things myself (even did a writeup on my site on how to setup a Software RAID array of external drives under Windows - yeesh!), I have to say that I can see some of the advantages to this concept.
However, given that I personally work with a lot of video, and have about 2 Terrabytes of storage used (with about 1 T of free space), and much of that data needs to be available "on demand" for the rest of the family to view, I'd be concerned about how differing transfer patterns would affect my monthly bill.
For myself, I ended up going with two Buffalo TeraStation Pro NAS 2.0 TB devices. They can be accessed using Windows, Linux or Mac OS - and support FTP as well. Since they are NAS devices, I was able to put them on a seperate network card so that the data is "isolated" from the outside world, and yet still available for playing back through the home Tivos. And the two units are small enough that I could carry them with me in an emergency (though not as easily as an id and password for S3 I would suspect).
Just my .02 worth.
Make sure you test jungle disk across platforms. Last time I tried data upload from a PPC Mac OS X couldn't be read by linux, nor the other way around. I know they patched this for intel mac to PPC mac, but am unsure of OS portability.
@Dossy Shiobara: Dreamhost is also ridiculously unreliable..
While you're at it, have a look at S3 Backup http://www.maluke.com/s3man/ it's a Windows app that does incremental backups and has compression and encryption functionality. Linux and Mac ports are not available at the moment but are planned (it's python and wx, so that's reasonable). I am the developer of this solution, if you have questions or suggestions, let me know.
I second the recommendation for Carbonite. It's already saved my ass once, and it just does it thing automagically.
"you cannot exceed 1 call per second OR send files > 40K"
I think you're supposed to read the *entire* license, and especially the section titled "Amazon Simple Storage Service", which, among other things, says that "The limitation of 1 call/per second/per IP address set forth in Section 1.A.2 above is not applicable to Your use of Amazon S3".
Have you seen bingodisk, http://www.bingodisk.com/, from joyent / textdrive. It's 200G of storage for $199pa hosted on the new Sun Thumpers, accessed via webdav. Don't have an account, but I'm thinking of running the maths, as you have, to see how it compares, to my TeraStation and S3. Unfortunately I don't have a Kill A Watt (and they're not available on amazon.co.uk) so I'd have to guess electricity.
I've written a whole blog series on using S3 for backup - see http://www.tunesafe.com for an index to the entries. It covers a lot of points made in these comments.
Basically Amazon's licence agreement is a mess and you shouldn't use S3 for anything you aren't willing to lose, at least until they sort out the legal stuff properly. Don't think that encryption will solve all the problems: in fact it leaves you open to yet more legal risks in the future.
S3 is certainly revolutionary and I'm using it now (a) to back up our Cardbox databases automatically every day - http://cardbox.wordpress.com/2006/07/11/amazon-s3-and-cardbox-2/ - and (b) to store full-resolution copies of every digital photograph I take, while keeping lo-res (480 x 360) copies on a searchable Cardbox database on my own local PC.
The database backups could be destroyed by Amazon on a whim, according to their agreement, but I also do weekly ones onto DVD. The photos aren't backed up and again Amazon could delete them or do anything else they liked to them, but at the end of the day a bit of bitrot is probably not such a bad thing.
One comment on a comment: the limit on the number of Amazon Web Services calls per second doesn't apply to S3. You can read/write as much and as often as you like.
I guess it depends on how many machines are being backed up to the backup server.
Perhaps the cheapest option is turn off the back up server when you don't need it.
This would totaly depend on how often your data is backed up (daily, hourly, weekly).
For my home stuff, I just have a removable hard drive. Does this trick. Very low budget. Although doesn't work as off site backup if my house burns down or gets robbed.
It seriously comes down to your needs.
I have a "backup server" at home that is running some mirrored disks in a small shuttle xpc. I use this to store all the photos/media/document that I need to access from my different systems and xbox media center.
History has taught me that this machine can and will choose to die at the worst possible time. I am now backing the critical stuff remotely with Backupright, http://www.backupright.com/ , I have it scheduled to run twice a day and it sends me an email when it fails.
Having remote backups has already paid off several times - regardless of what solution you go with, the benefits out weigh the costs.
What about tape drives? Ok, I know, I know, but there is some merit here, especially with the amounts of data being talked about here.
The Exabyte VXA2's can be got used for about $300 to $400 on eBay, and they store 80 GB natively (160 compressed, but I've never seen it). A SCSI card can be got for $75+. If you get 2 tapes - 1 for the full and 1 for incrementals, you're still only talking about $600 or so total cost. Plus, if you get an external unit, you don't even have to leave it powered up.
I've been using VXA's since I paid to have the data recovered of failed drives that were in a RAID 0+1 array where 2 of the 4 drives failed AT THE SAME TIME (yes, it does happen, regardless of how improbable it is!), and I've been quite happy with them.
I have my server set up with a VXA320 under Fedora Core 5 (which recognized the drive without any problems), and have been using star (s-tar, not "star"), and while it took me a week or so to get it working the way I wanted, I am quite happy with it now. If it spits the current tape out, I know it's full and needs another; Otherwise it just keeps incrementally backing up the folders I told it to. When I do a new full backup, I reuse the old tapes, as expected.
I recently got a working Seagate 25GB/50GB DAT drive off eBay for $8 (just had to clean it, as always for DAT's). Tapes are as low as $4. If you need to backup 10 or 20 GB, the only thing cheaper, but less flesible, is DVDR's (unless using RW's, but that's a whole different issue).
If you need to backup 1000's of GB it's one thing, but for 25 GB, 50 GB or even a couple of 100 GB, tapes can still work.
I'm just saying...
I wonder how the price would compare if instead of a full server configuration. If you used something like the Linksys NSLU2 you could just put your existing drives in USB enclosures or replace the server with on of the newwer external drives with built in NAS support.
I started using Amazon S3 to store some data too. i found it a bit difficult to get going for Mac OS 10. But now that I have it going, it's fairly simple.
And as this article implies... when you see what Amazon is charging you for the service, you will think they made a mistake. It costs me like PENNIES to use. I don't even have any idea how much data I have thrown out there yet. Why should I care when they're charging me such a reasonable amount.
Dang,I shouldn't be writing that! Some bean counter from Amazon will read this and raise the rates on us.
Also a former dreamhoster. Ever since their growing bandwidth sale/change their service and reliability have been terrible. They are way oversold and it shows. I had my simple gallery 1.0 of my son on there and it would take literally 3-5 minutes to load the index. Loads on my shell server rarely ever were below 100.
Did anyone consider the cost of air conditioning during the hot summer months. The air conditioner is a requirement, especially for a 5 drive system. Seems that A/C would put a big dent in a localhost system.
In response to the original backup solution with a P4 box: why get new, fast hardware just to copy data? Just rescue any old computer from the dumpster and stuff some big hard drives in it. If you use Linux, or another OS that doesn't depend so much on the BIOS to recognize hard drives, you don't have to deal with older BIOS' size limitations. Set the BIOS to turn the machine one every day before the backup runs if it supports it. If not, have the machine being backed up wake your backup box over the LAN. Have your scheduled backup script or program shut the computer down when the backup is done.
This way, the only things you need to actually spend money on are hard drives and a small amount of electricity, so it's not nearly as expensive as the original setup.
Personally, I back up anything irreplaceable to an old Pentium 133MHz box with 1 80GB and 1 160GB hard drives, running Debian Stable and rsback. It automatically turns itself on (thanks to a BIOS setting) every night at around 2AM, copies everything important to me (business stuff, school assignments, various projects, photos, etc.). I may buy a larger hard drive for it soon so I can also back up non-irreplaceable-but-a-pain-to-get-back stuff, like my music collection. Currently, less important stuff like that I back up to DVD-R, but only occasionally due to laziness.
Pros of this setup:
Cheap (no investment other than hard drives; very little electricity used)
Environmentally friendly (recycling an old computer, and again not using much electricity)
Locally available data
Can use the backup computer as a replacement fileserver/NAS if necessary
Fully automatic, so even busy/forgetful/lazy people can have a current backup
Cons:
No offsite backup included (unless you locate the computer offsite, of course)
May have to replace noisy fans with silent ones
May have to deal with flaky/unstable hardware (so test it well)
Another computer to maintain
Failures are not immediately apparent (so check it every so often, or have it send you emails every day upon completion or something)
If you need top flight data center backup, encryption, and all the other bells and whistles I got this in the email a couple of days ago, and it seems to be a pretty good deal. Here is the link...
http://www.netmass.com/infoworld_online_backup_special_offer.html
Or, you could buy 1 320GB external HD for $90 then make it external with an HD encasement $20. Then if you ever need more space, you can buy another external HD and hotswap them in the encasement.
Wow I just saved you over $1,500! Do i get a percentage?
Is it strictly a backup server, or also a fileserver?
I have a combo fileserver / backup server, but I would still have to keep it on all the time since I serve video/mp3 files off it.
Otherwise S3 sounds really cool. Will probably be using it eventually for some other online projects.
As a small web hosting company, I must advise everyone to be very careful trying to use a regular web hosting account for backup.
Much of the industry is in down-in-the-mud price war and one of the common techniques is to offer huge amounts of storage and bandwidth for ridiculous prices. (Some hosts still advertize completely "unlimited" disk storage).
The big "catch" is in the fine print know as the TOS (Terms of Service) or AUP (acceptable use policy). Buried in there are nice tidbits that basically say the site can only be used for actual html pages and storage of anything else (zip files, mp3, general binary files) is a violation and accounts can be terminated instantly with no access to the data.
The increased availability of WebDAV and other straightforward "net storage" and "net backup" programs is going to blow this wide open as many of this small hosts start seeing lots of customers actually trying to use all the capacity of their $2/month hosting accounts.
Also, beware of Carbonite and other "Web 2.0" startups offering backup services. Many of them are just front-ends to Amazon S3 throwing the dice and hoping you won't actually use a lot of storage.
In the case of Carbonite, their TOS basically says they can, at any time, declare your use of their service (in either amount of data stored, or bandwidth activity) to be "excessive" and shut you down. The problem is, they do not specify what constitutes abuse, so they can re-define it anytime they want. This week, you are allow 100GB of storage for $5/month, next week, 100GB of storage is declared to be "abusive" and they shut you down.
I personally view S3 as innovative and if I'm going to use any of these services, I would use S3 before anyone else.
I still feel it should only be a BACKUP not an alternative to onsite/local storage so the cost isn't saving the cost of large drives/local backup systems, but acts as a fail-safe in the case of fire, flood, e.g. where your computer and data might be completely lost or destroyed.
> switching to S3 could save me $587 over five years!
Dude, you're trying to save $100 a year?!
> My home server costs me a minimum of $13 every four weeks just to leave powered on and idling, or $170 per year.
Turn off the server when you're not using it.
Or get a server with a RAM drive or CompactFlash drive or something that will powerdown the hard drives and use very low power when not in use. That's more of a geek challenge than using S3.
Instead of trying to save $100 a year, how about trying to make a $100 a year? Put $2000 in a money market or savings account making 5% interest and there's your new $100. And you still get to keep and use the $2000.
It's one thing to get geeky and play "tech", but Amazon's S3 doesn't really make sense financially for me, either personally or for an online storage business. I've run and rerun and rerun the numbers and it just doesn't make sense.
Check out Duplicity (http://www.nongnu.org/duplicity/). It will encrypt your data with GnuPGP and do differential backups to Amazon's S3.
Hey Jeremy,
S3's turned out to be a lifesaver for us, and we've learned a lot through the process, so I'm happy to give you a hand or let you pick my brain if that'd be helpful.
Lunch, maybe?
Don
Did any of you ding dongs read the FAQ? Only 5 gigs! So much for the idea of backing up the entire drive... I have about 5 GB in my trash can never mind my hard drives...
I use ibackup at http://www.ibackup.com Never a problem. Very cheap.
http://rsync.net/ is $1.80 GB per month with no transfer fees and as the name suggests, is designed for mounting and use with existing backup tools.
I'm curious as to why anyone would use these backup services to store photos, when for ~$2/mo Flickr gives you unlimited storage (though a 2gb/calendar month limit on uploads).
I have 5 sets of data.
Photos: ~5gb and growing
Movies: ~200gb
Music: ~20gb
Personal documents: ~2gb
Applications: ~10gb
How much data do you have that needs to be protected against site failures? I have 7gb. And a DVD burner. And a safe.
Anyone know what kind of hardware S3 is sitting on? How do we know their infrastructure is much better than ours?
For small volumes of data, the new free X-drive is an option. 5GB of free storage, can be configured as a drive in Windows, automatic backup software (including scheduled), direct download from the Web.
Leaving out my pictures (stored on Flickr), music (iPod) and video (not worth backing up), my personal files come to ~2.5gb.
@KieranMullen: Did you read the faq? 5GB per object, unlimited objects.
I just bought the Kill-a-Watt on Amazon after reading about it here. I run a bunch of machines at home as well all doing different things. My curiousity kills me. I need to know what the power consumption is now. I told my buddy about it as well and his immediate reaction was "Must. Have. It."
Well.. I think you should not have bought such a pricey server when you bought it a few years back.. I picked up some rock-solid Compaq server hardware from someone on Ebay for $50 (w/ local pickup) and found someone in my neighborhood throwing away a small waist high equipment rack which is now holding my Proliant 6400R. I then picked up a used Compaq DLT tape drive (35/70Gb) for $40 and recently picked up more hot-swappable drives from someone on Craigslist (146Gb SCSI U320 drives for $50/ea).
Needless to say, my entire server is using redundant power supplies, has quad Xeon’s (Pentium III @ 550Mhz) and all file systems are RAID-5 and hot-swappable.. This machine is rock solid running FC3 and I’ll be keeping this tradition going (using older IT hardware) for the future as well.. For doing backups, you do NOT need anything faster than 1Ghz since it’s not really going to speed up any “processing”. I use my server for remote X11 development in Smalltalk, it’s also my mail host, database host (for Pgsql, and MySQL), Apache, etc.. The nice thing about this machine is that parts for it are really cheap and if you can’t find what you want/need, you can buy a 2nd machine for parts for pennies on the dollar! I’ve got someone in my local area (on Craigslist) selling two newer Proliant servers, one Sun server and 4-5 rack-mount drive enclosures for $120.. My point is that you just don’t need to fork over big $$ for this sort of hardware.
Also, if you’re concerned about automation, try out one of Compaq’s DLT tape libraries!
One of these : http://tinyurl.com/gf22s
For instance will backup ~1Tb since it has twin DLT drives and 3 cartridges each holding 5 DLT tapes.. Nothing like automation.. That drive has a buy-it-now price of $300.. Sure it’s a giant piece of hardware, but you’ll have something virtually foolproof for the next 5 years easily.. Anyway, food for thought!
Aren't you missing the most important part? Using S3 you need to BUILD an app to support your backup. If you are planning on using an existing, 3rd party backup app built on S3, you haven't explained that.
I agree with the suggestion for tape backups. It eliminates problems like corrupted files being backed up over your only backup (you can rotate tapes and archive master data images), reduces electrical costs (tapes don't need power when sitting in a drawer or safe deposit box), and you control physical security of your data.
The biggest issue I have with something like S3 is that all the stuff I want to backup is at home, where Comcast only gives me about 35 kB/sec upload speed. Sending gigabytes of RAW photos up to S3 would be pretty painful.
So, basically:
1. DreamHost.com Rocks
2. Amazon S3 has a future
3. jungledisk.com is nice
4. If your paranoid you backup in your basement every week
5. Every other link is shameless advertising
WOOT!
With no useful guarantee this strikes me as very risky.
Also, as data rates are still horribly poor, it is too slow on recovery for all but the most casusal users.
How long will it take to rcover 250GB from S3?
The best solution?
Hard disks, in removable hotswap trays.
Fedex or post office to an alternate location.
250GB hard disk in hotswap tray is now under $150 ( well under, but let's be conservative)
3 of those in rotation between / on 2 sites by courier.
Yes, I admit, user intervention is required.
But, when the chips are down, a backup disk or two with instantly available data ar but a few hours away, or of the one spare disk is on sote, just seconds away.
Sneakernet still wins this contest, hands down.
I agree with you. Amazon server has many benefits and is much better than other servers. Good luck with the backup work.
One thing is you're using an inefficient and costly server. An Infrant ReadyNAS storage server is around $1000 with 1 TB of disks, and draws about 55W. If you amortize it over 5 years, that's about $1390 for 5 years.
Go Daddy only costs 6 cents per GB per month with no transfer fees, so unlike S3, you won't have to pay $25 to restore from your backup. 5 years would cost $719.40, less than half S3.
This is pretty low-tech, but to really drop your power consumption you could consider putting the server on a timer. The server would only need to be up for the time that your backups are running. After that, it could easily shut itself down. A timer costs less than $10. Assuming your backups take 2 hours you could save 91% of your electric costs by using a timer.
I could see how it could be less of a hassle but the entire "cost of power" argument is a bit tricky.
- What about people who have got rented apartments that include all energy costs (including electricity) in the rent. They think of electricity as a "free service". No? (don't know about the US but its quite common here in The Netherlands)
- Also you could reduce on the "120 watts of power with the CPU idle and disks spinning" cost with wakeup on LAN?
- muneeb
So I come from the world of "sensor networks" where we are always concerned with reducing the power consumption of things as much as possible. Tiny sensors sleep most of the time and wakeup for very little timeslots to see if they have traffic to send or receive. We are working on an ACTUAL wakeup radio that takes very very small energy and you can wakeup the real radio by using the wakeup one (wakeup radios have been there in theory but i havent really seen actual ones)
So from the top of my head if the real cost of hosting a home server is electricity there are many ways to cut down on that (assuming that you are not fetching data from your homeserver all the time - which is a reasonlable assumption i think). I am not saying that i have a solution, I am just saying that i dont see this as a hard problem! .. theoretically your homeserver sleeps (consuming the absolute bare minimum electricity) and wakesup when there is an actual request for data retrieval - you could have a small wakeup LAN card that takes very small energy and its only job is to power up the system when it gets the data request then the other (actual) LAN card could be used to transfer data.
I work for a major financial firm and have the best storage servers / nas fillers etc at my disposal, but I simply do not need it. A 250GB external HD works fine for me. Any pictures / home movies get backed up on a DVD ( hopefully soon on blue ray). Now, discussing the price of electricity, well how much is it going to cost you to run your pc to upload 5gb's or more of data? Also, there is a question of security. Who knows who has access to these servers? I mean lets face it, what do you really need a separate server for for your home use? But thats just my two cents..everyone has their own needs.
2 more things:
1 - One of my considerations is who is backing these online storage facilities. At the moment, Yahoo owns flickr, so I am pretty confident flickr will be around in at least the medium term to store my photos. Similarly Amazon is backing S3 so that gives me a fair degree of comfort. Who is backing the other services such as dreamhost or carbonite?
Backup is all about peace of mind, which a flaky startup doesn't really give me.
2 - Most, (but not all) of my music & video files are backups of physical copies that I own. Having a look at Amazon's licence terms it says I warrant to them that I own the copyright to whatever I upload. If I breach this term then I have 60 days to rectify this - I presume by taking down the content. Does anyone disagree with this analysis?
(I think there is also a separate legal argument about whether backing up a cd that you own to mp3 is technically legal but lets leave that one aside)
Hi Jeremy
Excellent article, but I am anxiously waiting for you next article on how you did it.
I'm a newbie and it seems to me that I have to write programs/scripts using SOAP or REST just to put and get my data to and from Amazon S3? Or use free programs like s3backup.exe (http://www.maluke.com/s3man/#screenshot) to do my backup? s3backup.exe does not have a scheduler to schedule the backup -- and can we trust third party software with our secret key?
And I have a few linux and freebsd servers with data which I want to back up to S3. Right now, we are using "rsync" to backup the data to our backup servers (freebsd/linux). And we also have archive servers (freebsd) which we use Subversion (svn) to store our old files -- and we are thinking of using Amazon S3 for this -- but after reading up, I don't think Amazon S3 is suitable. Neither is it simple. It looks more for web developers?
Thank you.
Take a look at IDrive-E. This application allows you to backup 2GB of data for free! Simply download and install it and start backing up all your important and favorite data.
Now read this. IDrive-E gives you hands-free automated backups' of files and folders. You can back up 2GB of data free with no limits on upload size. You can backup any type of files. IDrive-E does incremental backups that transfer only portions of file that have been modified or changed since the last backup. Just select data, click on a button to back them up immediately. Or schedule the backups for a particular time and frequency and have a nice sleep.
IDrive-E retains 30 versions of backed up data. Each backup creates a new backup set that is identified by the date and time. You can restore up to 30 prior versions, including the most recent version of your data files. IDrive-E lays strong emphasis on encryption of data. All data transfer operations are protected with 256-bit AES encryption. IDrive-E encrypts data twice before it is sent from the user's computer. The data is first encrypted with a private encryption key stored on the user's computer. Then 128-bit SSL (the same protocol your web browser uses for secure web transactions) is used to transmit all data to and from our servers.
The application presents two interfaces for users to work with. IDrive-E Classic interface offers a very simple and user-friendly interface like Windows explorer to backup and restore files and folders, schedule your backup for a future data and time, exclude files and folders from backups and delete files and folders in the IDrive-E account. You have multiple options for backing up data: backups can be performed immediately or they can be scheduled for a date or time of your choice.
The IDrive-E Explorer view is meant for restoring files and folders and is not for backups. You can browse your IDrive-E account contents, restore files and folders with a simple drag-and-drop or copy-and-paste operation, view history of files, drag-and-drop or copy files to local drive (restore file versions) and search for files and folders backed up in your IDrive-E account.
2 words, Tilana Reserve. Simply set which files and/or folders you would like to protect and it does the rest. Simply put, I highly recommend this service to anyone who uses a computer. Their website is www.tilana.com
I tried www.disksave.com and they ended up inexpensive and worked out great.
My home storage server is an old P3-based laptop (which didn't cost me anything, I had it lying around anyway), with an attached 320GB USB 2.0 disk drive. When idle (USB drive spun down, internal drive spinning), and the networking up (screen disabled completely), the setup consumes 12W of power.
S3 will have to get pretty cheap to beat that; plus I have it available all the time... So the real advantage of something like S3 is off-site backup...
In addition, this machine also doubles as my mail server, proxy server, openvpn server and streaming audio/video server.
I documented my backup scheme on my personal site
http://michaelhorowitz.com/bbackup.html
I make both local and off-site backups at the same time, keep multiple copies of files that change often and both compresses and encrypt the backup copies. For the most part all the user (me) has to do is a quick review of the daily log to make sure everything worked. 30 seconds a day.
I run the backups once a day, your mileage may vary. This scheme did involve purchasing commercial software ($50) but I'm not married to the program (WinZip) and document its pros and cons.
The scheme is only for your most important files, not for huge videos or audio collections.
As for paranoid, the program doing the encryption is totally unrelated to the company providing the off-site storage space. And, there is no software constantly running in the background watching for file updates. That sort of thing scares me.
How will the new S3 pricing affect your backup system? Are you planning on staying with S3?
mozy is best of all - for 4.99 a month, you get unlimited back up.
I think s3 is a terrible solution for backing up- mozy has software as well, so it's all automatic..
onsite is still best, imho..
This is great...I've been doing my research on using Amazon S3 and I'm definitely sold. I'm planning on building a site that's going to stream lots of video - my lesbian gather membership site, and it still seems like I'll come out ahead. Thanks for sharing your thoughts on the issue.
For online backup, and regular file sharing, look at http://www.myotherdrive.com. They are the best game in town for free (5GB) and paid options (up to 200GB). Offer hyperlink (URL) access to your content, and bulk uploads. Check it out.
There are other value content hosting and delivery companies. Check out ValueCDN with rates as low as $0.10 per gig. Nice review thought!
Before deciding which service to use, there is a very important question you should ask yourself: what are you really looking for - remote storage, content delivery, or both.
What I see from my experience is that most people treat Amazon S3 more as a content delivery service. While this is not inherently wrong, S3 is primarily designed as a storage service.
The point is, if you are concerned about content delivery (offloading HTTP requests for static content from your webserver) and NOT about storage, there are much better services especially designed for that.
SteadyOffload.com provides an innovative, subtle and convenient way to offload static content. The whole mechanism there is quite different from Amazon S3. Instead of permanently uploading your files to a third-party host, their cachebot crawls your site and mirrors the content in a temporary cache on their servers. Content remains stored on your server while it is being delivered from the SteadyOffload cache. The URL of the cached object on their server is dynamically generated at page loading time, very scrambled and is changing often, so you don’t have to worry about hotlinking. This means that there is an almost non-existent chance that the cached content gets exposed outside of your web application.
It’s definitely worth trying because it’s not a storage service like S3 but exactly a service for offloading static content.
Watch that:
http://video.google.com/videoplay?docid=-8193919167634099306 (the video shows integration with WordPress, but it is integrable with any other webpage)
http://www.steadyoffload.com/
http://codex.wordpress.org/WordPress_Optimization/Offloading
Cost of bandwidth comes under $0.2 per GB - affordable, efficient and convenient. Looks like a startup but lures me very much. Definitely simpler and safer than Amazon S3.
I have a home backup server, it's an old 800MHz machine with a Debian Stable installation, it's plenty fast enough for rsync.
I've set the 'Wake on Alarm' settings in the BIOS to make it to power on at 16:00 every day, and cron runs a backup script when it starts up. The script uses rsync to backup my server and my desktop (which is usually on at 16:00), and ends with 'shutdown -h 1'. Any errors are emailed to me.
I was considering using Amazon S3 instead, but I don't think I'll bother. The biggest risk to losing my data is theft, so I think I'll just take my backup server to my parents' house to solve that one.
I am trying to use Amazon S3 for backup along with S3 Backup from maluke. com. The sofware is impressive. The problem I found is all the document created/modified dates are reset to the date of backup. Has anyone used S3 Backup and found the same problem? Or found something better to use with S3? thanks.
I found this article - even though its old I thought I'd add my 2cents.
I use an offsite backup server (similar to S3 - but much much cheaper and flexible) - its good for me because I can host my websites and stuff too. Regardless the solution is the same.
I use rsync on my mac to upload only changed files, and compress the hell out of the transmission. My dsl is slow, but once the bulk of the data is there the incremental updates only take 15-20minutes tops.
And I definately agree - using these offsite servers are better for peace of mind, upgrading at scale and having more space when needed.
I'm leading to online storage as well. One question, though...did you scrap your home server? If you did then your math is right. However if you still have your home server then you are just adding to your overhead by adding another monthly bill.
P.S. I also use kill-a-watt to measure my home office power consumption. What a great little gadget.
My problem with a lot of the above hardware tape based solutions is that tape will over time lose it's data, your tape drive _will_ become obsolete, so in 10 years time you won't be able to find a replacement, and your backup tapes will go from a useful data store to a collection of interestingly shaped paperweights.
You will most likely have to use the same backup programme to restore your information as you backed it up on (and find a machine with a SCSI interface and Windows 98 and a 5.25" drive to reload the original programme.).
I've had several tape drives go down in the past and their replacements haven't read tapes made with the originals (head alignment going out)..
Online backup seems to be the way to go, and the larger the company the better, so Amazon should be pretty safe. As they're charging you per gigabyte, they'll be able to afford to upgrade their own infrastructure as necessary, rather than the companies who offer as much as you can eat for a set amount.
Your S3 costs assume you never have to download your data. Depending on the number of drive failures over 5 years, your S3 cost will increase.
Amazon's price may go down over the next 5 years or go up. Is the current price for S3 storage just an introductory offer to get you hooked?
Jon mentioned using http:www.myotherdrive.com for online backup and I would like to point new features that let this site standout among others:
- Unattended Backup
- Encryption
- Copy To My Drive
- Grab File
Unattended Backup
The service copies new and changed files for me on a scheduled basis, protecting my data from loss. Each night when the backup completes, I receive an email summary report.
Encryption
Not comfortable storing all of your data on a online site? Then use the encryption feature.
Before my files are sent to www.myotherdrive.com, the files are encrypted using advanced AES 128-bit encryption. This means that when the servers receive my file, the contents are completely encrypted impossible for anyone on the Internet to read. When downloading encrypted files, I enter my password, and the files are decrypted on my machine before being written out. It's that simple.
Copy To My Drive
Have a file that needs to be shared amongst many? With this feature, one person uploads, and the rest can "Copy To My Drive" - effectively sending the file from the uploader's account directly to my account, bypassing any upload or download. I control the group that can share the file(s).
Grab File
When I see an interesting video or other file that I wish to move straight to your MyOtherDrive account, without downloading and then uploading, I use Grab File. Enter a URL, select a file name and folder, and it will be downloaded from the URL straight to your account.
I hope this info is helpful.
I use Amazon S3 - pretty cost effecient if you backup under 10TB. if you compare annual cost of storing 10TB of data on
Amazon S3 with the cost of external hard drive of the same size, i think you might want to consider the latter. I have my own freeware tool for Amazon S3 if you want to check it out http://cloudberrylab.com/
Another very useful tool: S3fm, a free online Amazon S3 file manager. 100% Ajax, runs directly from amazon s3, secure and convenient.
http://s3.amazonaws.com/s3fm/index.html
And if you want to have some fun while you are there be sure to check out built-in document viewer, photo slide show and media player.. :)
Actually you should not count the cost for motherboard, CPU, case, RAM, and so on since you still need to pay for that hardware anyway...
Sure makes sense switching to S3. A minor price increase with far better results. Nicely done.
I just realized that this post is 3 years old. I wonder how much the prices have fluctuated...
-Jack @ http://mozy.com
Hi,
I use D-Link 323 NAS with two 1GB SATA (RAID) via FTP anywhere in the world. My NAS goes into sleep mode after a certain period of time and there is no dedicated machine running and consuming power apart from one blue LED so this is not a problem for me and cost me less than Amazon.
I would only consider Amazon if:
1) I didn't have to keep and pay for my internet connection to be able to back-up data only (my parents are using it and have to keep it anyway)
2) My bandwidth was either limited or too slow
3) Have small files to do rare backups
4) Didn't already have purchased a UPS for power-cuts
In contrary on the above Amazon's cloud gives you probably more consistent bandwidth from any place in the world especially if you are traveling a lot, availability might be just about the same i.e. I've had to call home to "push the button" after a longer power cut once in over 2 years.
So if you are a photographer for instance, backing up a lot of NEF files, costs of using Amazon will probably outweigh having a similar set up to mine in a longer run. All come down to one's needs.
I was searching for Amazons' S3 and came across your blog and have been reading along the blog as well as the comments. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. although this post is old look at the comments by users. Still active and will be. Nice information.
A couple things, the $587 savings over 5 years is pretty insignificant. Just like one of the first comments, if you lose internet access, you lose access to your data. My recommendation would be to have both, noted that it is more expensive, but price should not subjugate data availability. Also look at the fact that internet backup is roughly 100 times slower than network backup. A gigabyte may be only 20 cents but it will take at least an hour to upload one, and in this day and age, we have several gigabytes of data to deal with.
i read a few comments ...
now i am confused ...
should i use Amazon or what ... ???
Wow, I just saw this dates back to 2006 and amazingly enought the information is still valid. The ones still
in operation are the ones you want - longevity and stability are always a good thing where backups are concerned.
I thought AWS S-3 was what all the real Pro's use but now I'm not sure either. I'm with subcorpus on that now.
The hardware that is being sold with Windows Home Server is extremely power efficient. Less than $5 a month to run, guaranteed. For $400, the Home Server is a no brainer, in my opinion. Perhaps, you're not realizing how little power this thing requires. An Atom 1.6 Processor and 120 Watt Power Supply with 2GB will do just fine.
My company's been examining the benefits of using Amazon's service as well. I was actually criticizing the web admin, until I looked over the pricing.
I still like running a personal server at home though.
I'm currently searching for a home file server and automatic backup solution.
Summarizing the comments so far:
1. Online storage offers peace of mind but likely only if you also back up locally.
2. A home file server is still needed for home systems with more then one computer, so the cost savings is likely only for the extra backup disks required.
3. Electricity costs are not as significant with the newer hardware which can go to sleep/hibernate and wakeOnLan when needed.
4. There are other online storage options that may be cheaper even much cheaper.
5. The ability to actually restore an entire system may end up being prohibitively slow on the server side so this needs to be carefully examined.
My own take is that online storage only offers a more convenient offsite storage IF the backup can be fully automated including failure notices. You'll still want a fast onsite fully automated solution. So far however, I've not been able to find this solution... Best solution so far? See http://productivegeek.com/forums/topic/windows-home-server-backup-to-lan
I have been working through the same issues and have written it up as a backup diary here:
http://martinaharris.com/tag/off-site/
In summary cloud storage like S3 was expensive and had no integrated facility. i.e its just storage. So I looked at the very cheap memopal. This was awful. Totally unreliable so stay well clear of them.
Then I moved on to crashplan. Which is awsome. Reliable, quick and fully functional. Not much more expensive than memopal, and much cheaper than S3. You can even backup to other friends machines. Good stuff. See the articles for detail though. And good luck.
Hello! I was searching for Amazons' S3 and came across your blog and have been reading along the blog as well as the comments. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. although this post is old look at the comments by users. Still active and will be. Nice information.
Thanks and regards.
I saw someone comment about they still like to have a personal server at home. Why?
Backups to the web all have one fatal flaw. Speed. You can back your data up and that happening very slowly is not a problem usually. Where The speed problem shows up is the "oh crap, I didn't think about how long this was going to take" moment when you realize that your data is safe, but your still screwed if you have a business to run because you can't get it back in less than a lifetime.
I agree with the slow speed issue - Backing up my approx 300Gig of data over the internet will take me weeks. Over the network it takes a day, and then rsync sorts me out nicely with tiny frequent incrementals.
Also, if you use a hulking old Pentium, it will run at 120w/h. If you build a 1.5Tb miniITX system, with atom dual core, it comes in at about 30-40W/h, and only costs you £300 ish.
That's a big saving on time and money compared to S3.
I have a Synology DS409+ and very happy with it. Low power usage, easy fast backups and allot of features. But the problem using your own server for backups is what happens when your house burns down or all your stuff get stolen, etc. With off-site backups like S3 the chances of losing your data is extremely slim. They broke into my buddies house a week ago and stole his pc and backup server ... 10 years of data gone forever!
Another option might keep another backup of all your stuff on a portable backup drive and then carry it with you or leave it in your car and update it once a week.
Or maybe leave it at work.
Just some options i'm considering ... main thing being I need a backup off-site.
