Machine-learning optimization for computing networks
A machine-learning optimization of a plurality of networks is provided. The machine-learning optimization includes interconnecting an online platform providing a machine learning module, a core network of computers deploying novel software, and a plurality of Internet network service providers. The platform collects, via the software, performance data of the Internet networks, which the machine learning module utilizes to enhance performance and reduce the latency therein networks by taking into account thousands of real-time and historic latency and bandwidth metrics. Thereby the software continually selects an optimal path through the plurality of Internet networks.
1. A computer-implemented method of optimizing data flow through a plurality of networks, comprising:
providing an online platform providing a machine learning module;
interconnecting the online platform and a plurality of network providers via at least one computer having software for causing the at least one computer to perform the following process steps:
(a) acquiring a list of open shortest path first (OSPF) neighbors of the plurality of network providers;
(b) acquiring a list of label distribution paths (LDP) neighbors;
(c) executing for each OSPF neighbor a bandwidth test for collecting latency information thereof;
(d) prompting the machine learning module to predict multiple path outcomes through the plurality of networks and calculate a confidence score of each path outcome based on latency information for associated OSPF neighbors; and
(e) making routing decisions for each OSPF neighbor based on its respective confidence score,
wherein if the confidence score is higher than 60 percent, then 80 percent of the collected latency information of step (c) is used by the software to determine a reliability value of each respective OSPF neighbor.
2. The method of claim 1 , wherein the latency information includes packet loss patterns through each OSPF neighbor, wherein each instance of packet loss is time stamped so that the software identifies sequential loss or maximum throughput of the each OSPF neighbor.
3. The method of claim 1 , wherein the confidence score is based upon historical information.
4. The method of claim 1 , further comprising the step of inspecting the confidence score via the software to tell the machine learning module whether it is correct or not.
5. The method of claim 1 , wherein if the confidence score is less than 60 percent, then 50 percent of the collected latency information of step (c) is used by the software to determine a reliability value of each respective OSPF neighbor.
6. The method of claim 1 , wherein OSPF neighbors comprises a list of objects including a router ID, a LSA types and an advertisement.
7. A computer-implemented method of optimizing data flow through a plurality of networks, comprising:
providing an online platform providing a machine learning module;
interconnecting the online platform and a plurality of network providers via at least one computer having software for causing the at least one computer to perform the following process steps:
(f) acquiring a list of open shortest path first (OSPF) neighbors of the plurality of network providers;
(g) acquiring a list of label distribution paths (LDP) neighbors;
(h) executing for each OSPF neighbor a bandwidth test for collecting latency information thereof;
(i) prompting the machine learning module to predict multiple path outcomes through the plurality of networks and calculate a confidence score of each path outcome based on latency information for associated OSPF neighbors; and
(j) making routing decisions for each OSPF neighbor based on its respective confidence score,
wherein if the confidence score is less than 60 percent, then 50 percent of the collected latency information of step (c) is used by the software to determine a reliability value of each respective OSPF neighbor.
8. The method of claim 7 , wherein the latency information includes packet loss patterns through each OSPF neighbor, wherein each instance of packet loss is time stamped so that the software identifies sequential loss or maximum throughput of the each OSPF neighbor.
9. The method of claim 7 , wherein the confidence score is based upon historical information.
10. The method of claim 7 , further comprising the step of inspecting the confidence score via the software to tell the machine learning module whether it is correct or not.
11. The method of claim 7 , wherein OSPF neighbors comprises a list of objects including a router ID, a LSA types and an advertisement.