Predicting Failure

 < Day Day Up > 



Recognizing the need for something more broadly functional than RAID, the server industry began developing methods to predict when a hard drive was going to fail so as to do something proactive instead of waiting until after the disk crashes. Most drive failures do not occur suddenly; rather, a slow degradation in performance can be detected, eventually culminating in total failure (Exhibit 3). By monitoring the status of drive performance indicators, then, it is possible to predict when a disk will fail and provide adequate time to safeguard the data, run diagnostics, and replace the drive if necessary before it completely fails. Disk monitoring also opens the door to being able to schedule disk replacement for slow times, perhaps during the middle of the night for servers that have to be up 24 hours a day. It is rather like watching the gas gauge on a car. The driver knows approximately how much fuel is left and can schedule a trip to the gas station whenever it is most convenient. Now imagine instead having to rely on the engine going dead before becoming aware that the fuel tank is empty. Who knows what kind of cargo the vehicle might be carrying, how far it will be from a gas station when the fuel runs out, or what time of the night it might be?

Exhibit 3: Sudden versus Gradual Failure (Black line shows non-predictable failure; performance remains good until a sudden failure occurs. Gray line shows predictable failure — steady performance decline over time.)

start example

click to expand

end example

The first firm to address this issue was IBM (Armonk, New York), which developed a technology called predictive failure analysis (PFA). PFA monitors elements within a disk drive that are invisible to the end user but which eventually lead to failure. For example, as the spindle bearings start to wear out, the increasing friction leads to a rise in temperature. An upsurge in the number of sectors that need to be remapped also indicates a problem with the disk surface.



 < Day Day Up > 



Server Disk Management in a Windows Enviornment
Server Disk Management in a Windows Enviornment
ISBN: N/A
EAN: N/A
Year: 2003
Pages: 197

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net