About the MTBF prediction problem

[ericzgh168]

For the networks product,how can we get a higher MTBF?More than 250,000 hours(prediction numbers) is for Cisco products,such as switching exchange,router,which is provided on the CISCO web.Most of the components or units are series except for the power. As I know,the 300,000 hours is a very difficult target for the prediction of the power supply.So,how can we get that high number for a system which using so many components?Is that true?

[tarik]

Are you asking about how did Cisco achieve a high MTBF for their networks products? Or how could MTBF of a network product be improved?

[ericzgh168]

Yes,I want to know how Cisco achieves the high MTBF for their product.I have ever caculated for the power supply.The value is offen 150,000 hours for PC,or 300,000hours for networks power.The communication system has so many components,inculding some high failure rate components ,such as FET,RAM,ROM and fans. I think the 250,000 hours or 300,000 hours for some CISCO systems is misson impossible. I don't believe it.

[tarik]

Some of the things that can be done to achieve high system reliability, in addition to improving component’s reliability, is creating redundancies and many backups to mitigate for components that have high failure rates in addition to performing preventive maintenances. Also note that many times companies report a high MTBF or a high Availability because they assume that all preventive maintenances are performed regularly also many times the time to arrive to the site or to obtain parts is not included into the time-to-repair calculations that goes into the system availability calculation.

I am unable to answer that question as I have never worked for Cisco. I suggest you contact them and have them address your question and tell you how they calculate the MTBF, what assumption they make and what sort of warranty they give and what are the terms of that warranty policy. Companies are usually resistant to tell how they actually perform their tests and make their analysis.

[DlCraggs]

MTBF is a rather poor measure of quality. If the failures are exponentially distributed, then about 63.2% of the items have failed at the MTBF point. The reliability is only 36.8%! NOT GOOD!

For warranty, you should be considering where the early failures occur. Perhaps the 1%, 5% or 10% failure points. This corresponds to the 99%, 95% and 90% reliability.

Another complicating factor is confidence. What confidence level does Cisco apply to the MTBF estimate? In many industries, 90% confidence is used.

[tarik]

all valid points!