Comment Re:Never underestimate the bandwidth (Score 1) 267
Regardless, NAS/SAN is a very viable solution for backups. Why duplicate your full dataset for every full backup you retain and exponentially increase your tape count, when you can do one full backup ever and only keep the incremental changes from that point on? Why pay someone to pick up tapes for offsite data protection, pay them to store it, pay them to retrieve it, and wait for as long is it takes for someone to get a tape that MIGHT have the specific data you need restored, when you can replicate the few gigs of incremental change weekly on a several hundred terabyte dataset that is actually new or changed to a site not just in the same city (as is usually the case of tapes) but to another state or even another country? And have the entire history of backups online and available in a few seconds and from a fail-over location? If you have a city-wide disaster and you've got your tapes just down the street in a "vault", you're probably screwed. If it's 1000 miles away your business/research group/agency can recover.
Every point the article makes about why tape is an advantage is easily disqualified by anyone with basic understanding of the technologies. I mean seriously, the article actually argues that a tape is superior because if a tape snaps it can be spliced back together and you only lose "a few hundred megabytes", whereas if you lose a TB disk drive you lose the whole TB. If you're too fucking stupid to use striping then you deserve to lose the data.
It further argues that you dont have to pay to power tapes. Sure, true. You just have to pay to ship it around, pay for it in terms of floorspace to store it, and do so exponentially more so than disk (see dedup argument above). But the article aslo proves perfect ignorance of NAS/SAN tech allowing for very low cost idle disks. So the disks remain spun down until/unless the data on them is accessed. There is minimal cooling/power costs associated unless you're accessing the data regularly over the whole of the storage, and if you're doing that you've further negated any professed benefit from tapes.
And the article also gives no consideration for technology refresh for "long term" data storage, which shows an amazing lack of foresight. If all your data is on LTO tape, for instance, what happens in 5 years? Do you copy all that data from old tape tech to new tape tech? Or do you keep all or some subset of the old tape hardware online and available so that you can still access the old tapes? If the latter, do you start paying more and more for the support on that hardware? Or do you risk failures that you may or may not be able to find replacement parts to repair? If it's the former, is it easier to copy the pb's of data that the LHC project is describing from one tape tech to another? Or from one disk storage tray to another?
The more I consider the article the more it pisses me off because it's so flat out false. There might be a very small set of scenarios in which the article's view is viable, but they are by far the exception rather than the norm.