Controlling a working condition of electronic devices

Patent metadata

Field	Value
Publication number	US-11419247-B2
Application number	US-202016829513-A
Country	US
Kind code	B2
Filing date	Mar 25, 2020
Priority date	Mar 25, 2020
Publication date	Aug 16, 2022
Grant date	Aug 16, 2022

Abstract

An approach for controlling a working condition of electronic devices via controlling a climate parameter. The approach comprises measuring a climate parameter distribution; feeding the climate parameter distribution to a climate control system; obtaining operational data from each of the electronic devices; feeding the operational data into the climate control system to determine control actions; obtaining activity data about external activities; providing a machine learning algorithm trained with past activity data; feeding new activity data to the machine learning algorithm; feeding the prediction output to the climate control system; generating a current climate map; generating a target climate map; and generating a rearrangement plan with rearrangement steps for the electronic devices.

First claim

What is claimed is:

1. A method for controlling a working condition of electronic devices via controlling a climate parameter, the method comprising:
measuring a climate parameter distribution by an array of climate sensors arranged in vicinity of an array of the electronic devices;
feeding the climate parameter distribution to a climate control system and controlling by the climate control system the climate parameter by providing a controllable fluid stream of a cooling fluid towards the electronic devices;
obtaining operational data from the electronic devices, representing working conditions based on workloads, and based on the climate parameter affecting the electronic devices;
feeding the operational data into the climate control system to determine control actions to be executed to control the working condition of the electronic devices by controlling the climate parameter;
obtaining activity data about external activities influencing the workload processed by the electronic devices;
providing a machine learning algorithm trained with past activity data, thereby having learned to derive from the activity data an expectable workload for the electronic devices being expectable in a predetermined time window after the external activities;
feeding new activity data to the machine learning algorithm, and obtaining from the machine learning algorithm a prediction output regarding the expectable workload for the electronic devices, in a predetermined time window;
feeding the prediction output to the climate control system as additional input for controlling the climate parameter in the predetermined time window;
generating a current climate map, the current climate map reflecting a climate profile across the array of the electronic devices, obtained via the controlled climate parameters;
generating a target climate map, the target climate map being obtainable via rearranging the electronic devices, wherein the target climate map represents a workload and climate conditions matching the current climate map at a reduced power consumption over a predetermined second time window; and
generating a rearrangement plan, comprising rearrangement steps for the electronic devices, required when starting from the current climate map to arrive at the target climate map.

2. The method according to claim 1 , wherein the climate parameter represents an operating temperature of an electronic device.

3. The method according to claim 1 , wherein the controllable fluid stream of the cooling fluid is an air flow, controllable by variable supply fluid dampers, in particular supply air dampers, supplying the cooling fluid to the array of electronic devices, in particular air inlets in raised ground floor plates.

4. The method according to claim 1 , wherein the rearrangement steps are performed automatically, in particular by a robot.

5. The method according to claim 1 , wherein the machine learning algorithm is based on a deep learning algorithm, in particular an adaptive deep learning algorithm.

6. The method according to claim 1 , wherein the external activities comprise at least one of social media entries, news media, weather information, energy market data, stock exchange data.

7. The method according to claim 1 , wherein the machine learning algorithm comprises an architecture for natural language processing for evaluating the external activities, in particular a Deep Belief Network.

8. The method according to claim 1 , wherein the machine learning algorithm is trained with intrinsic factors, comprising at least one of a geographic location, a date, a timestamp, a sensor measured value, a sensor position, a fluid damper position, a configuration management, incident/problem management data, a vendor specification, an access management data, a house automation data or an industry type.

9. The method according to claim 1 , wherein the climate parameter distribution is measured by providing climate sensors at different locations across the array of electronic devices, in particular in a mesh-like array of climate sensors.

10. The method according to claim 9 , wherein the climate sensors are positioned on a front side of the array of electronic devices.

11. The method according to claim 1 , wherein the climate sensors are at least one of a temperature sensor, a humidity sensor, an air flow sensor, an acoustic sensor, a vibration sensor, a gas sensor or a particle sensor.

12. The method according to claim 1 , wherein pre-training the machine learning algorithm comprises a Restricted Boltzmann Machine.

13. The method according to claim 1 , wherein fine-tuning the machine learning algorithm comprises a back-propagation algorithm.

14. The method according to claim 1 , further comprising:
fetching intrinsic data and external data of data centers in one or more industries;
deducing at least one of intrinsic indicators or external indicators;
determining a time-lag between the indicators and climate changes;
correlating the indicators with the climate changes and applying the time-lag;
pre-training a prediction model associated with a Deep Belief Network using a Restricted Boltzmann Machine;
using the prediction model as initial input for a feed-forward neural network;
fine-tuning the prediction model using a back-propagation algorithm; and
verifying the prediction model by applying a test dataset.

15. The method according to claim 1 , further comprising:
determining at least one of anomalies in the working condition or failures of the electronic devices by evaluating a measured climate parameter distribution.

16. The method according to claim 1 , wherein a measured climate parameter distribution is continuously compared with manufacturer specifications of the electronic devices.

17. The method according to claim 1 , further comprising:
determining at least one of failure probabilities or a temperature prediction for an array of electronic devices using one or a multitude of pre-trained Deep Belief Networks.

18. The method according to claim 1 , wherein at least one of fluid dampers are controlled pro-actively or an alarm is raised, depending on at least one of failure probabilities, a temperature prediction, a target temperature or an alarm threshold.

19. A computer program product for controlling a working condition of electronic devices via controlling a climate parameter, the computer program product comprising:
a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer system to cause the computer system to perform a method comprising:
measuring climate parameters by an array of climate sensors arranged in vicinity of an array of the electronic devices;
feeding the climate parameters to a climate control system and controlling, by the climate control system, the climate parameters by providing a controllable fluid stream of a cooling fluid towards the electronic devices;
obtaining operational data from the electronic devices, representing working conditions based on workloads, and based on the climate parameters affecting the electronic devices;
feeding the operational data into the climate control system to determine control actions to be executed to control the working condition of the electronic devices by controlling the climate parameters;
obtaining activity data about external activities influencing the workload processed by the electronic devices;
providing a machine learning algorithm trained with past activity data, thereby having learned to derive

Assignees

• Kyndryl Inc

Inventors

• Egger Michael
• Zenerino Toby
• Eggenberger Christian
• Kauke Joerg
• Bindi Dario

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• H05K7/20836 Primary

  Thermal management, e.g. server temperature control · CPC title