CrashPlan, and how I use it
I've been advising others on backup solutions recently, and so I thought I'd share the way I use CrashPlan to keep my machines backed up. It's not, as far as I can tell, how most would use it, but it provides me with the flexibility I like.
I've got a pair of Windows desktops, and a pair of Ubuntu boxes, one of each at home and in my office. CrashPlan, which is Java based, is available for all platforms, so it's installed on all 4 machines, which are linked under one account. The two linux boxes hold my data on RAID arrays (no, RAID is not a substitute for backup, but it doesn't hurt).
There are two things I don't like about cloud backup services - the potential cost of bandwidth to push files to the cloud, and the ease with which you could in the event of a catastrophe, get your data back.
The solution to the first problem is that I pay my ISP an extra $10 a month for unlimited uploads. This allows me the chance to transfer as much data as I like. I'm not making the very best use of that potential, as I'm not sending backups to the cloud, but it certainly cuts the cost of doing backup, or removes the possibility of accidently using all your allocated bandwidth.
To get around the second problem, I use a peer <--> peer vs client <--> cloud model for my backups. One of the most elegant features of Crashplan is the ability to backup between installed machines, instead of just to the cloud. Each of my desktops backs up to each of my linux boxes, one local and one remote ; each server backs up locally to its RAID array, and to the other server. This means that I've got automated local and offsite backups, both under my control.
To avoid the onerous initial bandwidth requirements, Crashplan allows you to seed a backup by making a local one to removable media, and then moving it via 'sneakernet' to the remote location. After it's been attached at the remote end, it can then be used as a backup target, with only changes being transfered, moving forward.
The only issues that I've got with this setup are limitations in the way the Crashplan software is set up. I want to run backups on different schedules and with different file sets (local vs remote) which requires a subscription to the Plus version of the software, to unlock the functionality. I'm paying the base / machine cost for this, per year, of $25USD/machine, and get 10GB of cloud storage per machine that I don't use. I understand, subsequently, that, for the number of machines I've got, I could have done better by choosing an 'unlimited' family plan.
There's been recent heated discussion of Crashplan's switch to a local AU based data centre, with increases in pricing of the order of 25%. It's one of those situations where Crashplan can't win; some customers want a local data store, which is costing the company considerably more to run - Australia's not a cheap place to do business. Some customers don't care where their data ends up; no-one gets a choice. Add to this the lack of multi-year plans, and locals are a bit shirty.
According to the current pricing structure, I would still consider a move to the $165AUD / year family unlimited plan; there are a few more machines I could add to the backup, and I could push to the cloud on a lesser schedule as another level of redundancy. This would also give me access to my backups via the Android (or iOS) client.