APPLICATION STATE PREDICTION USING COMPONENT STATE
Described systems and techniques enable prediction of a state of an application at a future time, with high levels of accuracy and specificity. Accordingly, operators may be provided with sufficient warning to avert poor user experiences. Unsupervised machine learning techniques may be used to characterize current states of applications and underlying components in a standardized manner. The resulting data effectively provides labelled training data that may then be used by supervised machine learning algorithms to build state prediction models. Resulting state prediction models may then be deployed and used to predict an application state of an application at a specified future time.
1 - 20 . (canceled)
21 . A computer program product, the computer program product being tangibly embodied on a non-transitory computer-readable storage medium and comprising instructions that, when executed by at least one computing device, are configured to cause the at least one computing device to:
receive a data stream of performance metrics characterizing a technology landscape, the technology landscape including at least one application provided by at least one component, the at least one application being associated with a service model that describes causal relationships between a plurality of components, including the at least one component, and the at least one application;
detect a temporal pattern of component states of the at least one component, based on the performance metrics; and
predict a future application state of the at least one application, based on the temporal pattern of component states.
22 . The computer program product of claim 21 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
receive the data stream of performance metrics including a set of Known Performance Indicators (KPIs) characterizing the at least one component and the at least one application;
reduce the set of KPIs by applying a Principal Component Analysis (PCA) to the set of KPIs and thereby obtain a reduced set of KPIs; and
determine a component state of the at least one component and an application state of the at least one application at corresponding timestamps, based on the reduced set of KPIs.
23 . The computer program product of claim 21 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
receive the data stream of performance metrics including reports of user interactions with the at least one application; and
determine, based on the reports, a component state of the at least one component and an application state of the at least one application at a corresponding timestamp of each report of the reports.
24 . The computer program product of claim 21 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
train an unsupervised machine learning algorithm to construct at least one anomaly detection model;
receive the data stream of performance metrics; and
determine component states of the at least one component and application states of the at least one application at corresponding timestamps, using the anomaly detection model.
25 . The computer program product of claim 24 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
apply each component state of the component states, and each application state of the application states, as a label of each corresponding timestamp, to obtain labelled training data;
train a supervised machine learning algorithm using the labelled training data, to obtain at least one state prediction model; and
predict the future application state, using the at least one state prediction model to classify the temporal patterns of component states.
26 . The computer program product of claim 24 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
classify the component states and application states with respect to anomaly thresholds, using the at least one anomaly detection model; and
generate an alert when the one or more of the component states or the application states crosses one of the anomaly thresholds.
27 . The computer program product of claim 21 , wherein the service model includes causal connections between all of a plurality of components that includes the at least one component and the at least one application.
28 . The computer program product of claim 27 , wherein the instructions, when executed, are further configured to cause the at least one computing device to: predict the future application state of the at least one application, based on the temporal pattern of component states and on the causal relationships.
29 . The computer program product of claim 21 , wherein the instructions, when executed, are further configured to cause the at least one computing device to: predict the future application state of the at least one application, including predicting a bounded time window during which the future application state will occur.
30 . The computer program product of claim 21 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:
detect a temporal pattern of application states of the at least one application, based on the performance metrics; and
predict the future application state of the at least one application, based on the temporal pattern of application states.
31 . A computer-implemented method, the method comprising:
receiving a data stream of performance metrics characterizing a technology landscape, the technology landscape including at least one application provided by at least one component, the at least one application being associated with a service model that describes causal relationships between a plurality of components, including the at least one component, and the at least one application;
detecting a temporal pattern of component states of the at least one component, based on the performance metrics; and
predicting a future application state of the at least one application, based on the temporal pattern of component states.
32 . The method of claim 31 , further comprising:
training an unsupervised machine learning algorithm to construct at least one anomaly detection model; and
receiving the data stream of performance metrics and determining component states of the at least one component and application states of the at least one application at corresponding timestamps, using the anomaly detection model.
33 . The method of claim 32 , further comprising:
applying each component state of the component states, and each application state of the application states, as a label of each corresponding timestamp, to obtain labelled training data;
training a supervised machine learning algorithm using the labelled training data, to obtain at least one state prediction model; and
predicting the future application state, using the at least one state prediction model to classify the temporal patterns of component states.
34 . The method of claim 32 , further comprising:
classifying the component states and application states with respect to anomaly thresholds, using the at least one anomaly detection model; and
generating an alert when the one or more of the component states or the application states crosses one of the anomaly thresholds.
35 . The method of claim 31 , wherein the service model includes causal connections between all of a plurality of components that includes the at least one component and the at least one application.
36 . The method of claim 31 , further comprising: predicting the future application state of the at least one application, based on the temporal pattern of component states and on the causal relationships.
37 . The method of claim 31 , further comprising: predicting the future application state of the at least one application, including predicting a bounded time window during which the future application state will occur.
38 . A system comprising:
at least one memory including instructions; and
at least one processor that is operably coupled to the at least one memory and that is arranged and configured to execute instructions that, when executed, cause the at least one processor to
receive a data stream of performance metrics characterizing a technology landscape, the technology landscape including at least one application provided by at least one component, the at least one application being associated with a service model that describes causal relationships between a plurality of components, including the at least one component, and the at least one application;
detect a temporal pattern of component states of the at least one component, based on the performance metrics; and
predict a future application state of the at least one application, based on the temporal pattern of component states.
39 . The system of claim 38 , wherein the system is further configured to:
train an unsupervised machine learning algorithm to construct at least one anomaly detection model; and
receive the data stream of performance metrics and determine component states of the at least one component and application states of the at least one application at corresponding timestamps, using the anomaly detection model.
40 . The system of claim 39 , wherein the system is further configured to:
apply each component state of the component states, and each application state of the application states, as a label of each corresponding timestamp, to obtain labelled training data;
train a supervised machine learning algorithm using the labelled training data, to obtain at least one state prediction model; and
predict the future application state, using the at least one state prediction model to classify the temporal patterns of component states.