Mean time between failures

He showed me all these data sheets.

They all had 50,000 hour MTBF in bold with a box around it and stuff, and, and, like, all these things, from itty bitty fans that fit on a single component up to great big blowers that fit on a great big box.

They all were exactly 50K.

And I'm like, "Well, that's amazing that you engineered them all to be the same." He goes, "What do you mean? That's just the thing on a spread- on a data sheet.

But some of that is pre...

Part of it is...

And I got this the other day from a, a comment, was, why is MTBF and MTTF and all this stuff and fit rates so prevalent? And, uh, my immediate response was, because it's so easy to get what appears to be a good answer, and then you don't have to worry about it.

Except reality doesn't care what makes you feel good.

Let's get lots of information in one shot and then give the results.

Because they are optimized for high throughput, you know, the way to kind of look at mainframe success is through this very interesting term called MTBF, which is mean time between failure.

It is a term that captures how long has a computer been operating before there is a failure.

And can you guess that for mainframes, that time MTBF is often measured in decades.

As in, if you commission a mainframe today, it may not go down for the next 10 years, 20 years or 30 years because it is engineered for reliability and throughput.

Back then in the 1960s and 70s, IBM was largely importing, refurbishing second-hand computers from developed markets.

For the senior viewers, you must also monitor your reliability metrics.

MTBF or meantime between failures tells you how reliable your hardware is.

The higher the number, the better.

MTTR, or mean time to repair, tells you how fast your team is.

So, for example, you have like machines that behave in a very, very predictable way.

And when manufacturer makes a machine like a computer, they put like MTBF rating, mean time between failure.

So you look at that number and then you look at that how many machines you have.

And it gives you like actually pretty decent numbers.

Mm-hmm.

But yeah, in terms of what does it take to remove a safety driver, it is, it's the safety case, and it's, it's proving out, uh, the mate- mean time between failures, the MTBF.

Um, and, and then there's also a, a, a, a very rigorous systems engineering aspect of this.

And so systems-

Yeah.

What's the MTBF on an LGA 1951? We'll have to ask Intel.

Yeah.

These are the questions that Anand isn't around to ask anymore, so it's up to us to-

Hey, hey, uh, Francois in the chats, uh, LGA-

He's there

... 1951 MTBF.

He is there.

Just give me a number.

And, you know, something close to your heart, MTBF.

You know how that's been misused over decades.

And still being misused.

And they're still poisoning.

And I still see it on LinkedIn.

How to effectively use MTBF and all this other stuff.

And I just go, guys, you know, and again, all the...

And part of that is to, at the start, is to reduce the uncertainty.

So where is it are we going? What direction are we going? And from our perspective, is this product, does this product need to be as good as our last product? Does it need to be as good or worse or good as better as our competitors' products? Do we have a specific goal? Hopefully not MTBF, but let's say it's, You know, 98% reliable over five years in a home environment.

And I'm thinking of my inkjet printer sitting over there that just finished printing.

You know, with a small home office kind of load on it and use on it.

We were talking about reliability metrics in different industries and different sectors.

And we talked about consumer products where warranty reliability is important, electronic components where they have this crazy reliability metric called FITS and the defense sector as well, which is madly obsessed with the MTBF.

And we realized that there's actually a lot more to talk about.

And I think two of the things that we want to talk about in terms of reliability metrics in this Reliability Metrics Part 2 podcast is medical devices and software.