Cognitive service request construction

Patent metadata

Field	Value
Publication number	US-11487604-B2
Application number	US-201715859552-A
Country	US
Kind code	B2
Filing date	Dec 31, 2017
Priority date	Apr 23, 2017
Publication date	Nov 1, 2022
Grant date	Nov 1, 2022

Abstract

An embodiment includes a method for use in managing a system comprising one or more computers, each computer comprising at least one hardware processor coupled to at least one memory. The method comprises a computer-implemented manager: detecting that the system is in an unhealthy state; determining a set of one or more possible actions to remedy the unhealthy state of the system; selecting at least one action of the set of one or more possible actions; and constructing a service request implementing the selected at least one action; wherein at least one of the detecting, determining, selecting, and constructing is based at least in part on applying a reinforcement learning algorithm.

First claim

What is claimed is:

1. A method for use in managing a system comprising one or more computers, each computer comprising at least one hardware processor coupled to at least one memory, the method comprising a computer-implemented manager:
detecting that the system is at risk of shutdown due to memory usage exceeding a threshold, based at least in part on a real-time assessment of an operating context of the system and a cognitive state of a user;
determining, using a reinforcement learning algorithm of the computer-implemented manager, a set of possible actions to remedy the risk of shutdown of the system due to memory usage exceeding the threshold, the set of possible actions including adding memory, adding a file system, adding a processor, and deleting log files;
identifying at least one resource, comprising a node and at least one corresponding application, relevant to the detected risk of shutdown of the system;
displaying at least one form template corresponding to the at least one identified resource, wherein the form template is selected, created and/or modified based on learning which content items and configuration parameters are best to display for a specific cohort, based at least in part on the real-time assessment of an operating context of the system and the cognitive state of the user;
identifying one or more possibly affected properties of the at least one relevant resource;
based on the identified one or more properties, displaying at least one content item within the form template;
selecting at least one action of the set of possible actions;
wherein selecting at least one action of the set of one or more possible actions comprises:
displaying one or more content items associated with the at least one action; and
obtaining one or more selections by the user of respective configuration parameters for the one or more content items associated with the at least one action;

constructing a service request implementing the selected at least one action to remedy the risk of shutdown of the system, wherein the service request is constructed based on a user interface displaying the form template, and wherein constructing the service request comprises populating the form template with one or more configuration parameters corresponding to the one or more content items; and
based at least in part on the constructed service request, implementing the selected at least one action to remedy the risk of shutdown of the system, by carrying out, on the computer system, at least a corresponding one of adding the memory, adding the file system, adding the processor, and deleting the log files;
wherein:
determining the set of one or more possible actions comprises computing one or more risk values each corresponding to each respective one of the set of possible actions, wherein one or more reward signals of the reinforcement learning algorithm are inversely proportional to the one or more risk values; and

selecting at least one action comprises selecting the at least one action based at least in part on the computed one or more risk values corresponding to respective ones of the set of possible actions, by selecting a next highest risk action above a lowest risk action when a lowest risk action has a hardware precondition which cannot be met.

2. The method of claim 1 , wherein computing the one or more risk values comprises a Markov decision problem (MDP) in which the probability is based only on a current state and action rather than on any prior states or actions.

3. The method of claim 2 , wherein computing the one or more risk values comprises a partially observable Markov decision problem (POMDP) in which the current state is unknown.

4. The method of claim 2 , wherein computing the one or more risk values comprises:
maintaining a library of discovered paths between the current state and the at least another state; and
exploring one or more additional paths, not in the library, between the current state and the at least another state.

5. The method of claim 4 , wherein the library of discovered paths is constructed at least in part using training data.

6. The method of claim 1 , wherein the reinforcement learning algorithm does not require training data or user feedback.

7. The method of claim 1 , wherein constructing the service request comprises at least one of:
deciding at least one of whether and when the service request should be issued;
conceptualizing and parameterizing the service request based at least in part on at least one of a topology and an architecture associated with the system; and
determining at least one service request relationship comprising at least one pre-condition and at least one post-condition.

8. The method of claim 1 , wherein the reinforcement learning algorithm utilizes:
data-driven dialog management;
correlating observed events; and
analyzing open and closed tickets.

9. The method of claim 1 , wherein constructing the service request comprises automated user interface reconfiguration based at least in part on a real-time assessment of an operating context of the system and a cognitive state of a user.

10. The method of claim 1 , further comprising, based at least in part on the constructed service request, implementing the selected at least one action to remedy the risk of shutdown of the system, by carrying out, on the computer system, at least a corresponding one of adding the memory, adding the file system, adding the processor, and deleting the log files.

11. The method of claim 1 , wherein computing the one or more risk values each corresponding to each respective one of the set of possible actions comprises applying:
R (θ,δ)= θ L (θ,δ( X ))=∫ χ L (θ,δ( X )) dP θ ( X )

where:
θ is a fixed state of nature;
X is a vector of observations stochastically drawn from a population;
δ is a decision rule over which the risk function is calculated;
L is a loss function;
E θ is an expectation over all population values of X; and
dP θ is a probability measure over an event space of X, parametrized by θ.

Assignees

• IBM

Inventors

• Kalia Anup
• Hwang Jinho
• Vukovic Maja
• Wu Frederick Y

Classifications

• G06N7/01

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

• G06F11/079 Primary

  Root cause analysis, i.e. error or fault diagnosis (in a hardware test environment G06F11/22; in a software test environment G06F11/36) · CPC title

• G06F11/0751

  Error or fault detection not based on redundancy (power supply failures G06F1/30; network fault management H04L41/06) · CPC title

• G06F11/073

  in a memory management context, e.g. virtual memory or cache management (memory management G06F12/00; testing of static memory units G11C29/00) · CPC title

• G06N20/00

  Machine learning · CPC title