Calculation of MTBF (Mean Time Between Failure)

[Banfa]

OK here's the problem, I have 1000 Network switches in a network. The switches are all rated as 1 failure in 300,000 hours.

After how many hours should I expect 1 of them to fail?

The other people here are saying 300 hours but I do not think that it is as simple as 300,000 / 1000.

This calculation does not seem right to me as it assumes that all the switches are mutually dependent as far as MTBF is concerned.

I'm saying 1500 hours for a 99.3% chance that at least 1 switch has failed.

[Ganon11]

Each individual switch is expected to fail once within 300,000 hours?

Then the only thing you can predict is that, after 300,000 hours, all of your switches will have failed at least once.

But I don't know - I hated probability.

[epifora]

If all or lot of switch had been broken without given mtbf time, you should be true mtbf was misscalculated.

That is given 300.000 hours means, that the switch work 300.000 hours in normal condition. (Normal Condition: Determined electric power, environment conditions, air condition, nears other electronic environments, etc.)

If this condition is changing variantly, mtbf time is exacly change.

Good think that: your switches are been good mtbf time
One of the broken is going to be services and then solved the problem in warranty time.

I know some mtbf calculation software, but not easy to calculate electronic equipment of mtbf. This job done from your producer.

One of an online calculation method MTBF Calculator for to run an eye over.

[gersla]

Dear all.
How to get a real MTBF if we know, that we sold in a year 100 devices and 10 of them failed during this period. What is MTBF of the device?

[tharden3]

@ gersla: please don't post a question on a thread if the original question has not yet been answered. It is rude, and very annoying to the original person trying to get help.

Kind regards,
Tim

[kthprog]

The MTBF of a system of components which are not serially dependent on each other (one failure does not mean the system will fail) is equal to the inverse of the sum of the inverses of the MTBF of each component, in this case each switch.

so 1/MTBF = 1/MTBF of Switch 1 + 1/MTBF of switch 2 ...

So the result would be 1/MTBF = 1000*(1/300,000)

1/MTBF = 1000/300,000

1/MTBF = 1/300

MTBF = 300

It's just a coincidence that the MTBF is in fact 300. This is the correct formula to calculate the MTBF of a system, based on the components MTBF, though.

[SwissProgrammer]

You used the term MTBF, which I think you might mean "Mean Time Between Failure". Therefore, it looks to me like someone reported this in relation to an analysis which in itself was reported via a bell curve. If you are looking at a bell curve report I would say that 2% might fail immediately. I would suggest planning for a continuous potential of 2% failure.

I suggest keeping at minimum 20 switches in stock and ready to use as replacements continuously. Include lead order time to replenish those 20 switches in the total backup inventory to keep a minimum of 20. Add 20 to have 20 immediately after a potential full count of 20 are used all on the same day: Thus in inventory a minimum of 40 switches, plus the lead order time extras.


Back to the "I have 1000 Network switches in a network. The switches are all rated as 1 failure in 300,000 hours."

What this says to me is that (if economically reasonable) a total replacement of all switches should be scheduled at or within 90% of the 300,000 hours use. At about 270,000 hours use have a (well supported with in-place at that time material and in-place at that time labor) plan to replace all of the switches. That means even the ones that have been replaced; to start the process over.

As an option watch the error rate and when you see that it has reached some level that you decide (30% sounds fine for me) do a total replacement at that time rather than consideration of the supplier's MTBF report.