VM-to-VM traffic estimation in multi-tenant data centers
Methods and systems for network management include performing path regression to determine an end-to-end path across physical links for each data flow in a network. A per-flow utilization of each physical link in the network is estimated based on the determined end-to-end paths. A management action is performed in the network based on the estimated per-flow utilization.
1. A method for network management, comprising:
performing path regression to determine an end-to-end path across physical links for each data flow in a network;
reducing a number of flows by filtering the flows according to one or more rules;
estimating per-flow utilization of each physical link in the network based on the determined end-to-end paths; and
performing a management action in the network based on the estimated per-flow utilization,
wherein performing path regression comprises determining a portion of an end-to-end path based on forwarding table information and determining a remainder of the end-to-end path based on one or more inference rules.
2. The method of claim 1 , further comprising:
calculating expected link counts based on the determined end-to-end paths;
comparing the expected link counts to measured link counts to generate an error measurement; and
adjusting the determined end-to-end paths based on the error measurement.
3. The method of claim 2 , wherein calculating the expected link counts comprises:
representing the end-to-end paths as a path matrix;
representing the measured link counts as a measured link count vector;
determining an inferred flow performance vector based on the path matrix and the measured link count vector; and
determining an expected link count vector based on the path matrix and the inferred flow performance vector.
4. The method of claim 1 , further comprising clustering the flows into flow groups.
5. The method of claim 1 , wherein performing the network management function comprises one or more of rerouting a flow, changing a topology of the network, and changing a setting of a device on the network.
6. A method for network management, comprising:
performing path regression to determine an end-to-end path across physical links for each data flow in a network;
reducing a number of flows by filtering the flows according to one or more rules and clustering the filtered flows into flow groups;
estimating per-flow utilization of each physical link in the network based on the determined end-to-end paths; and
performing a management action in the network based on the estimated per-flow utilization,
wherein performing path regression comprises determining a portion of an end-to-end path based on forwarding table information and determining a remainder of the end-to-end path based on one or more inference rules.
7. The method of claim 6 , further comprising:
calculating expected link counts based on the determined end-to-end paths;
comparing the expected link counts to measured link counts to generate an error measurement; and
adjusting the determined end-to-end paths based on the error measurement.
8. The method of claim 7 , wherein calculating the expected link counts comprises:
representing the end-to-end paths as a path matrix;
representing the measured link counts as a measured link count vector;
determining an inferred flow performance vector based on the path matrix and the measured link count vector; and
determining an expected link count vector based on the path matrix and the inferred flow performance vector.
9. The method of claim 6 , wherein performing the network management function comprises one or more of rerouting a flow, changing a topology of the network, and changing a setting of a device on the network.
10. A system for network management, comprising:
a path regression module comprising a processor configured to determine an end-to-end path across physical links for each data flow in a network;
a filtering module configured to reduce a number of flows by filtering the flows according to one or more rules;
a flow utilization estimation module configured to estimate per-flow utilization of each physical link in the network based on the determined end-to-end paths; and
a network management module configured to perform a management action in the network based on the estimated per-flow utilization,
wherein the path regression module is further configured to determine a portion of an end-to-end path based on forwarding table information and to determine a remainder of the end-to-end path based on one or more inference rules.
11. The system of claim 10 , further comprising a data validation module configured to calculate expected link counts based on the determined end-to-end paths, to compare the expected link counts to measured link counts to generate an error measurement, and to adjust the determined end-to-end paths based on the error measurement.
12. The system of claim 11 , wherein the data validation module is further configured to represent the end-to-end paths as a path matrix, to represent the measured link counts as a measured link count vector, to determine an inferred flow performance vector based on the path matrix and the measured link count vector, and to determine an expected link count vector based on the path matrix and the inferred flow performance vector.
13. The system of claim 10 , further comprising a clustering module configured to cluster the flows into flow groups.
14. The system of claim 10 , wherein the network management function comprises one or more of rerouting a flow, changing a topology of the network, and changing a setting of a device on the network.