CALIBRATION OF RESOURCE SERVER POWER ALLOCATIONS WITHIN A DATA STORAGE DEVICE USING CONFORMAL PREDICTIONS
Methods and apparatus for power management in data storage devices are provided wherein conformal prediction is employed to determine correction terms for applying to power-per-processing event (P/PE) values. One such data storage device includes a non-volatile memory (NVM), a set of hardware processing engines, and a power sensor to detect a total power consumption of the set of hardware processing engines. A processor is configured to determine a P/PE value for each of the set of processing engines based on total power consumption measurements using a least squares procedure. A conformalization procedure is applied to sequences of P/PE values to calibrate the P/PE values by determining correction terms for applying to the P/PE values to provide guaranteed power prediction intervals. Delivery of power to the processing engines is then controlled based on the corrected P/PE event values in accordance with a power budget. On-line and off-line examples are provided.
1 . A data storage device, comprising:
a non-volatile memory (NVM);
a plurality of processing devices, each configured to perform a corresponding processing event to process NVM data;
a power sensor configured to measure a total power consumption of the plurality of processing devices; and
one or more processors configured, individually or collectively, to:
measure a power-per-sequence value for each of a plurality of sequences of processing events using the power sensor;
determine an estimated power-per-processing event value for each of the processing events;
determine, based on the measured power-per-sequence values and the estimated power-per-processing event values, a conformity score for each of the plurality of sequences of processing events;
determine sequence-based correction terms from the conformity scores;
adjust the power-per-processing event values using the sequence-based correction terms; and
control delivery of power to the plurality of processing devices based on the adjusted power-per-processing event values.
2 . The data storage device of claim 1 , wherein the one or more processors are configured to determine the conformity scores using conformal prediction (CP).
3 . The data storage device of claim 1 , wherein the one or more processors are configured to determine the estimated the power-per-processing event value for each of the processing events by being further configured to perform a least squares procedure on a plurality of measured total power consumption values.
4 . The data storage device of claim 3 , wherein the one or more processors are configured to determine the conformity score by being further configured to determine a value that represents a cumulative deviation of the estimated power-per-processing event values from the measured power consumption value for each of the sequences of processing events.
5 . The data storage device of claim 1 , wherein the one or more processors are configured to determine the sequence-based correction terms from the conformity scores by being further configured to conformalize a plurality of the conformity scores.
6 . The data storage device of claim 5 ,
wherein the one or more processors are configured to conformalize the plurality of the conformity scores by being further configured to determine a histogram of the plurality of the conformity scores; and
wherein the one or more processors are configured to determine the sequence-based correction terms from the histogram.
7 . The data storage device of claim 1 , wherein the one or more processors are further configured to adjust the power-per-processing event values by being further configured to multiply each of the power-per-processing event values by a portion of a corresponding sequence-based correction term.
8 . The data storage device of claim 1 , wherein the one or more processors are further configured to control the delivery of power using parameters comprising one or more of:
control parameters received from a host, parameters representative of an operating status of the data storage device, parameters representative of a priority level associated with the processing events, and parameters representative of a current processing workload of the data storage device.
9 . The data storage device of claim 8 , wherein the parameters representative of the operating status of the data storage device comprise one or more of: a power supply status, a power mode status, a device age, a total number of write-erase cycles performed, a time elapsed since data write, an overall device temperature, a device temperature at program, a device temperature at read, and a difference between the device temperature at program and the device temperature at read.
10 . The data storage device of claim 1 , wherein the one or more processors are further configured to determine the power-per-processing event value for each of the plurality of sequences of processing events by being further configured to determine a least-squares fit of measured total power consumption values to a matrix product of the power-per-processing event values and a list of corresponding active processing devices.
11 . The data storage device of claim 1 , wherein the one or more processors are further configured to:
determine corresponding duration values for each of the processing events;
determine, based on the adjusted power-per-processing event values and the corresponding duration values, an energy-per-processing event value for each of the processing events; and
control delivery of energy to the plurality of processing devices based on the energy-per-processing event values.
12 . A method for use by a data storage device comprising a non-volatile memory (NVM) and a plurality of processing devices configured to process NVM data, the method comprising:
measuring a power-per-sequence value for each of a plurality of sequences of processing events using a power sensor;
determining an estimated power-per-processing event value for each of the processing events;
determining, based on the measured power-per-sequence values and the estimated power-per-processing event values, a conformity score for each of the plurality of sequences of processing events;
determining sequence-based correction terms from the conformity scores;
adjusting the power-per-processing event values using the sequence-based correction terms; and
controlling delivery of power to the plurality of processing devices based on the adjusted power-per-processing event values.
13 . The method of claim 12 , wherein the conformity scores are determined using conformal prediction (CP).
14 . The method of claim 12 , wherein estimating the power-per-processing event values for each of the processing events comprises performing a least squares procedure on a plurality of measured total power consumption values.
15 . The method of claim 14 , wherein determining the conformity score comprises determining a value that represents a cumulative deviation of the estimated power-per-processing event values from the measured power consumption value for each of the sequences of processing events.
16 . The method of claim 12 , wherein determining the sequence-based correction terms from the conformity scores comprises conformalizing a plurality of the conformity scores.
17 . The method of claim 16 ,
wherein conformalizing the plurality of the conformity scores comprises determining a histogram of the plurality of the conformity scores; and
wherein the sequence-based correction terms are determined from the histogram.
18 . The method of claim 12 , wherein adjusting the power-per-processing event values comprises multiplying each of the power-per-processing event values by a portion of a corresponding sequence-based correction term.
19 . The method of claim 12 , wherein the delivery of power is controlled using parameters comprising one or more of: control parameters received from a host, parameters representative of an operating status of the data storage device, parameters representative of a priority level associated with the processing events, and parameters representative of a current processing workload of the data storage device.
20 . An apparatus for use in a data storage device, the apparatus comprising:
a non-volatile memory (NVM);
a plurality of processing devices configured to process NVM data;
means for measuring a power-per-sequence value for each of a plurality of sequences of processing events using a power sensor;
means for determining an estimated power-per-processing event value for each of the processing events;
means for determining, based on the measured power-per-sequence values and the estimated power-per-processing event values, a conformity score for each of the plurality of sequences of processing events;
means for determining sequence-based correction terms from the conformity scores;
means for adjusting the power-per-processing event values using the sequence-based correction terms; and
means for controlling delivery of power to the plurality of processing devices based on the adjusted power-per-processing event values.