Method and apparatus for reliable I/O performance anomaly detection in datacenter

摘要

A method and apparatus for reliable I/O performance anomaly detection. In one embodiment of the method, input/output (I/O) performance data values are stored in memory. A first performance data value is calculated as a function of a first plurality of the I/O performance data values stored in the memory. A first value based on the first performance data value is calculated. An I/O performance data value is compared to the first value. A message is generated in response to comparing the I/O performance value to the first value.

主权项

1. A method comprising:
storing input/output (I/O) performance data values in memory, wherein
each of the I/O performance data values comprise
an amount of time it took for a device to respond to a request from any one of a plurality of other devices, andthe performance of the device is measured by the amount of time; calculating a first performance data value as a function of a first plurality of the I/O performance data values stored in the memory; calculating a first value and a second value based on the first performance data value,
wherein
the first value is an upper band, andthe second value is a lower band; comparing a new I/O performance data value to the first value and the second value; and generating a message in response to comparing the new I/O performance value to the first value and the second value, wherein
the message includes a warning if the comparison indicates that the new I/O performance value exceeds the first value or falls below the second value.