Predictive failure analysis
| Industry | Information technology, Computer hardware |
|---|---|
| Founded | June 16, 1911 in Endicott, New York, United States |
| Products | Predictive Failure Analysis (PFA), Hard disk drives, Storage devices |
| Website | www |
Predictive Failure Analysis (PFA) refers to techniques aimed at anticipating potential failures in hardware or software components, allowing for preventive maintenance or corrective measures prior to actual malfunction.
The term was initially introduced by IBM as a proprietary method to monitor the probability of failure in hard disk drives. Today, it broadly encompasses technologies that evaluate the likelihood of impending failures in CPUs, memory modules, and input/output devices. For example, systems track patterns in corrected error rates to forecast hardware or memory issues and take proactive steps to avoid them.
Disks
In 1992, IBM debuted PFA with its 0662-S1x model—a 1052 MB Fast-Wide SCSI-2 disk spinning at 5400 rpm. This technology focuses on monitoring critical mechanical factors like the flying height of the read/write heads, comparing measurements against predefined limits to assess drive condition. When a potential failure is detected, an alert is sent to the disk controller.
The initial implementation had certain limitations:
- The output was binary—only indicating whether a failure warning was present or not to the host
- Communication was one-way, with alerts transmitted exclusively from the drive’s firmware
Subsequently, this method integrated with IntelliSafe technology, laying the groundwork for Self-Monitoring, Analysis, and Reporting Technology (SMART).
Processor and Memory
Elevated numbers of corrected errors identified by ECC can serve as indicators of forthcoming DIMM failures. Systems automatically retire faulty memory pages and CPU cache lines to minimize error impact. For instance, on Linux, the mcelog daemon handles automatic offlining of memory sections and processor cores that exhibit frequent correctable errors, thereby improving overall system stability.
Optical media
Failures due to degradation in optical media such as CD, DVD, and Blu-ray discs can be predicted by analyzing correctable data error rates using software tools like QpxTool or Nero DiscSpeed. This allows early identification of deteriorating media quality before data loss occurs. Nevertheless, support for error scanning varies between different optical drive manufacturers and models.