International Business Machines Corporation

Kavya Govindarajan

Chander Govindarajan

Praveen Jayachandran

Mudit Verma

Seep Goel

Priyanka Prakash Naik

Mudhakar Srivatsa

Harshit Kumar

Joshua M Rosenkranz

Pranita Sharad Dewan

Pranjal Gupta

Aishwariya Chakraborty

Metric management techniques in a multi-cluster computing environment are disclosed. For example, a method obtains a set of metrics collected from one or more computing clusters of a distributed computing environment, wherein the set of metrics are associated with one or more processes that are executable on the one or more computing clusters. The method computes one or more metric importance values for the set of metrics based on one or more processing criteria. The method sends the one or more metric importance values to at least one computing cluster of the one or more computing clusters to enable the at least one computing cluster to adapt a collection frequency of at least one metric of the set of metrics. Such adaptation, by way of example, can be based on the metric importance and resource availability.

Metric management in multi-computing cluster environment

A computer-implemented method, according to one approach, includes: receiving one or more streams of system log information from one or more microservices, and comparing the received system log information to compressed templates. One or more blocks of the system log information that correspond to one or more of the compressed templates are identified. All static text is removed from the identified blocks. Moreover, runtime variables from the identified blocks are inserted into respective ones of the compressed templates. The streams of system log information are further updated by replacing the identified blocks with the compressed templates having the runtime variables therein.

Lossless compression of system log information

Edge application deployment in a network is provided. The network includes a plurality of edge sites with edge computing infrastructure. Edge application deployment is performed, including deploying a pseudo application instance (pApp) of the edge application at each edge site of a first group of edge sites of the plurality of edge sites, and deploying a real application instance (rApp) of the edge application at each edge site of a second group of one or more edge sites of the plurality of edge sites. The pApp is a lightweight, application-specific instance of the rApp with less application functionality than the rApp. Further, the first group of edge sites is larger than the second group, and a user device interaction with the edge application is through a selected pApp of the first group of edge sites to an rApp of the second group.

Enhanced edge application deployment and processing in a network

An ensemble of autoencoder models can be trained using different seeds. The trained ensemble of autoencoder models can be run on new time series data to generate a prediction associated with the new time series data. The new time series data can include multiple dimensions per time step. Reconstruction errors can be determined for the prediction. Dimensions having highest reconstruction errors can be selected among the multiple dimensions based on a threshold. The prediction can be segmented based on bursts of the reconstruction errors over time, where temporal segments can be obtained. At least one common pattern including a set of dimensions among the selected dimensions across the temporal segments can be obtained to represent a failure fingerprint.

Localizing faults in multi-variate time series data

Growing community of inventors

Chennai, India

Kavya Govindarajan