Determination of endpoint device location for efficient analysis of network performance
View Patent ↗A network monitoring and analysis system is configured to conduct performance analysis on a network using a plurality of endpoint devices. Locations of respective endpoint devices for use in the performance analysis of the network are determined utilizing an algorithm which characterizes the network as a graph having edges and vertices and associates endpoint devices with particular ones of the vertices in an iterative manner based on impact of said vertices on achievement of a specified coverage criterion. The coverage criterion may comprise a links-covered criterion which specifies a percentage of a total number of links of the network that are covered by test communications. As another example, the coverage criterion may comprise a links-estimable criterion which specifies a percentage of a total number of links of the network for which performance is estimable from collected end-to-end path measurement data.
1. A method comprising:
selecting at least two endpoint devices for use in performance analysis;
generating test communications between a pair of the selected endpoint devices;
collecting end-to-end path measurement data utilizing the generated test communications; and
transforming the end-to-end path measurement data to produce a plurality of performance indicators;
wherein the selection of the endpoint devices is based on the impact of the transmission of test communications along a path between the pair of endpoint devices on the achievement of a coverage criterion; and
wherein the coverage criterion comprises a links-covered criterion which specifies a percentage of a total number of links of the network that are covered by the test communications.
2. The method of claim 1 wherein at least one of the endpoint devices comprises a network device configured to have test communication generation capability.
3. The method of claim 1 wherein at least one of the endpoint devices comprises a device that is coupled to a network device that does not have test communication generation capability.
4. The method of claim 1 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, wherein the vertices correspond to switching elements of the network and the edges correspond to connectivity between the switching elements of the network.
5. The method of claim 1 wherein the coverage criterion comprises a links-estimable criterion which specifies a percentage of a total number of links of the network for which performance is estimable from the end-to-end path measurement data.
6. The method of claim 1 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, and wherein the algorithm at a given iteration selects one or more of the vertices for association with one or more respective endpoint devices based on scores computed for the vertices, wherein after selection of the one or more vertices in the given iteration scores of the remaining vertices are recomputed for use in a subsequent iteration.
7. The method of claim 6 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, and wherein the score computed for a given one of the vertices is given by a number of uncovered edges on paths between the given vertex and vertices already selected for association with respective endpoint devices.
8. The method of claim 1 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, and wherein the graph comprises a plurality of trees defined by the edges and vertices, and wherein the leaf nodes of the trees are associated with respective ones of the endpoint devices.
9. The method of claim 6 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, and wherein the score computed for a given one of the vertices is given by a rank of the given vertex after sorting unselected vertices based on a number of trees in which the vertices appear and degree of the vertices.
10. The method of claim 6 wherein the endpoint devices are selected based on an algorithm which characterizes the network as a graph having edges and vertices, and treats the endpoint devices as vertices on the graph, and wherein if a selected one of the vertices comprises a non-leaf node of a tree of the graph, the tree is broken at the selected vertex into a number of separate trees.
11. The method of claim 1 wherein a given one of the test communications is directed between a first one of the endpoint devices and a second one of the endpoint devices.
12. The method of claim 1 further comprising repeating the generating, collecting and transforming steps for each of a plurality of time intervals.
13. The method of claim 1 wherein at least one of the performance indicators comprises a binary indicator, the binary indicator taking on a first value to indicate that a corresponding link is not associated with a performance problem, and taking on a second value to indicate that the corresponding link is associated with a performance problem.
14. A method comprising:
selecting at least two endpoint devices for use in performance analysis;
generating test communications between a pair of the selected endpoint devices;
collecting end-to-end path measurement data utilizing the generated test communications; and
transforming the end-to-end path measurement data to produce a plurality of performance indicators;
wherein the selection of the endpoint devices is based on the impact of the transmission of test communications along a path between the pair of endpoint devices on the achievement of a coverage criterion; and
wherein for a given time interval the collected end-to-end path measurement data is characterized by the equation:
Y=Ax
where y is a vector of end-to-end path measurements, A is a flow matrix defining a selected pattern, and x is a vector of network link-level performance indicators.
15. The method of claim 14 wherein the transforming step comprises utilizing y and A to solve the equation for x.