Toshiba spins out new NAS disk drive with its fastest transfer rate yet

Re: A million hours MTBF?

It's no different to Miele claiming they've tested their washing machines for the equivalent of 20 years use and using that to justify adding £200+ to the price compared to the nearest equivalent (e.g a Bosch) but only offering a 2 year warranty (5 years if you're lucky).

Yes, it is different.

It's different because a hard drive isn't like a washing machine. When your washing machine stops working you need to repair or replace your washing machine, and maybe one load of washing (and, if you're really unlucky, mop up a lot of water). When your hard drive dies you need to replace (repair is unlikely to be possible) the drive and ALL the data you've ever stored on it, not just the data you're reading or writing at the time.

That makes a difference because while it is clearly acceptable (in the sense that lots of people do it, and people still buy their products) to deliberately make a less-than-perfect washing machine, it is definitely not acceptable to deliberately make a less-than-pefect hard drive. I say deliberately because we all know that even hard drives do fail, and sooner or later that will happen to all of us, so we keep backups (don't we?) ... the point is that no hard drive maker would ever stay in business by making a hard drive that was deliberately shoddy and liable to fail sooner rather than later, while washing machine manufacturers do it all the time.

[Incidentally, my Miele washing machine HAS been running without fault, averaging about a wash per day, for 20 years, and I rather doubt that I'd be able to make a similar claim of any other brand. It was a bit more than £200 more expensive than the alternatives, though (I think we paid £900 at a time that the Deutschmark was at an all-time high against the pound, and we could have had a cheap brand for about £250 or a Bosch for about £600). My Miele fridge and dishwasher have been going for 27 years.]

I suggested that if they were claiming they have been tested for that much use they should offer at least a 10 year warranty because if their machines were as reliable as their marketing material claimed the longer warranty cost to them would be negligible. I was told they couldn't possibly afford to offer a 10 year warranty which tells you all you need to know about their claimed reliability.

I think it tells you all you need to know about the way people treat equipment when they think that someone else will carry the can when it fails. People would try it on.

The thing is MTBF is lies damned lies a statistical measure based on intensive use over a short space of time with the actual failures extrapolated into an indicative MTBF value when used at a normal level. It doesn't really take into account real usage patterns over many years with the associated degredation of the materials over time.

I would rather argue that MTBF doesn't mean what you apparently want it to mean. It's a measure of the likely number of failures occurring in a large sample of items over a period of time. It doesn't tell you anything useful about the likelihood of any particular item failing in that time, or of the likely time before a particular item fails.

You're right though, that MTBF is often misunderstood, and is often used misleadingly.

You could argue MTBF is useful in comparing reliability between different mechanisms as a drive with MTBF=200k hours implies it's 2x reliable than a drive with MTBF=100k hours but that assumes the testing methodology hasn't been massaged to inflate the figures is directly comparable.

I find one has to make assumptions of that kind when assessing the content of any sales-oriented "literature".