I bit the bullet: effective today, we are giving S3 a spin as our nextgen backup solution. The experience so far has been quite impressive: it took a couple of minutes to set up an account, download a couple of nice tools and start backing up stuff. I must admit that doing backups in an era of virtual machines and LVM snapshot is much easier than it used to be, and having a virtually unlimited backup media at your fingertips can really make the difference.
There are still a lot of i’s to dot and t’s to cross indeed, yet using S3 for backups seems to be very promising. My cost overguesstimate, including bandwith for incremental backups and the occasional restore, lands around €2/GiB/year, which is definitely not bad for an off-site, fully on line, secure, serviced and completely outsourced solution. Assuming we get past the apparent bandwith limitations (we’re having peaks of 3Mbps, but it takes a few parallel backup processes to get there), I see no reason not to provide every Sourcesense employee with an S3 account, having no more excuses for failing hard drives and no more rushes to the stationery cupboard to grab the last DVD left in the office.
True enough, it’s not like we have stellar requirements when it comes to backups: all we need is storage, and we’ll do the scripts. We don’t have massive amounts of data, we don’t have convoluted applications requiring cumbersome hot backup procedures and tols, and most definitely we don’t even see us getting there in short time, which means S3 fits the bill just fine. I wouldn’t suggest it to a bank or a telco just now, but I do have the feeling it could really be a viable solution for a lot of SMEs and/or vertical scenarios.
While we’re at it, let’s not forget the Open Source side of SaaS. The good news is that the sky is not falling, and there is still a lot of room for hacking: just look at the impressive list of available S3 tools. Building such an amazing plethora of software didn’t require any knowledge about the Amazon internals: we don’t need to know whether Amazon is using some custom Linux storage driver or just moving disks around using elves and hobbits, as there has never been a real need to know how a chipset is designed to hack around it. Open Source just needs APIs and documentation, access to information that is, to thrive: luckily enough, it seems there are very few Nvidias and Broadcoms in the SaaS world. We’re in for some serious fun!