Methods and computer program products for correlation analysis of network traffic in a network device
First Claim
Patent Images
1. A method of analyzing network traffic comprising:
- collecting data related to network traffic within a computing network;
generating a plurality of metrics associated with the data, the plurality of metrics including network element metrics and node-to-node contexts;
selecting a primary metric from the plurality of metrics associated with the network element metrics and the node-to-node contexts;
identifying a candidate set corresponding to a portion of the plurality of metrics;
generating correlation coefficients between the primary metric and the portion of the plurality of metrics that are members of the candidate set;
generating a hypothesis set based on the correlation coefficients, wherein the hypothesis set includes the portion of the plurality of metrics that are members of the candidate set that are causally related to the primary metric;
identifying other metrics of the plurality of metrics with a high correlation to the primary metric, wherein a positive correlation suggests that a higher load on one network element or context is indicative of a causal link to increased utilization of another network element or resource and a negative correlation suggests that one process or transaction is indicative of a causal link to inhibiting another process or transaction from obtaining adequate resources;
adding the other metrics of the plurality of metrics with the high correlation to the primary metric to the hypothesis set; and
terminating a process associated with at least one metric of the plurality of metrics indicated in the hypothesis set;
wherein at least one of selecting the primary metric, generating the correlation coefficients, and generating the hypothesis set is performed using at least one processor.
4 Assignments
0 Petitions
Accused Products
Abstract
Provided are methods and computer program products for monitoring the contents of network traffic in a network device and performing correlation analysis of collected performance metrics to help identify reasons for network performance issues. Methods for correlation analysis include selecting scope and network metric types to include in the correlation analysis to generate an educated candidate set for correlation analysis. The correlation analysis methods result in a hypothesis set that assist the operator in identifying transactions and infrastructure problems resulting in network performance degradation.
24 Citations
34 Claims
-
1. A method of analyzing network traffic comprising:
-
collecting data related to network traffic within a computing network; generating a plurality of metrics associated with the data, the plurality of metrics including network element metrics and node-to-node contexts; selecting a primary metric from the plurality of metrics associated with the network element metrics and the node-to-node contexts; identifying a candidate set corresponding to a portion of the plurality of metrics; generating correlation coefficients between the primary metric and the portion of the plurality of metrics that are members of the candidate set; generating a hypothesis set based on the correlation coefficients, wherein the hypothesis set includes the portion of the plurality of metrics that are members of the candidate set that are causally related to the primary metric; identifying other metrics of the plurality of metrics with a high correlation to the primary metric, wherein a positive correlation suggests that a higher load on one network element or context is indicative of a causal link to increased utilization of another network element or resource and a negative correlation suggests that one process or transaction is indicative of a causal link to inhibiting another process or transaction from obtaining adequate resources; adding the other metrics of the plurality of metrics with the high correlation to the primary metric to the hypothesis set; and terminating a process associated with at least one metric of the plurality of metrics indicated in the hypothesis set; wherein at least one of selecting the primary metric, generating the correlation coefficients, and generating the hypothesis set is performed using at least one processor. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28)
-
-
29. A method of analyzing network traffic comprising:
-
collecting data related to network traffic within a computing network; generating a plurality of metrics associated with the data, the plurality of metrics including network element metrics and node-to-node contexts; selecting a primary metric from the plurality of metrics associated with the network element metrics and the node-to-node contexts; generating correlation coefficients between the primary metric and a portion of a plurality of collected metrics associated with the network element metrics and the node-to-node contexts that are causally related to the primary metric; and generating a hypothesis set based on the correlation coefficients, wherein the hypothesis set includes the portion of the plurality of metrics that are causally related to the primary metric; identifying other metrics of the plurality of metrics with a high correlation to the primary metric, wherein the high correlation indicates that processes associated with the other metrics are consuming more resources than metrics of the plurality of metrics without the high correlation to the primary metric; adding the other metrics of the plurality of metrics with the high correlation to the primary metric to the hypothesis set; and terminating at least one process indicated by a metric of the plurality of metrics in the hypothesis set to thereby free resources for remaining metrics of the plurality of metrics; and wherein at least one of generating the correlation coefficients and generating the hypothesis set is performed using at least one processor. - View Dependent Claims (30, 31, 32, 33)
-
-
34. A computer readable storage hardware including computer-readable program code that when executed by a processor enable the processor to:
-
collect data related to network traffic within a computing network; generate a plurality of metrics associated with the data, the plurality of metrics including network element metrics and node-to-node contexts; select a primary metric from the plurality of metrics associated with the network element metrics and the node-to-node contexts; identify a candidate set corresponding to a portion of the plurality of metrics; generate correlation coefficients between the primary metric and the portion of the plurality of metrics that are members of the candidate set; generate a hypothesis set based on the correlation coefficients, wherein the hypothesis set includes the portion of the plurality of metrics that are members of the candidate set that are causally related to the primary metric; identify other metrics of the plurality of metrics with a high correlation to the primary metric, wherein a positive correlation suggests that a higher load on one network element or context is indicative of a causal link to increased utilization of another network element and a negative correlation suggests that one process is indicative of a causal link to inhibiting another process from obtaining adequate resources; add the other metrics of the plurality of metrics with the high correlation to the primary metric to the hypothesis set; and terminate a process associated with at least one metric of the plurality of metrics indicated in the hypothesis set.
-
Specification