AI-supported network techniques
Examples of the present disclosure relate to an AI-supported CDN. In examples, a data processing engine processes log data of a CDN node according to a model to identify an issue. An issue indication is provided to a solution generation engine, which generates a set of solutions to automatically resolve the issue. The set of solutions is provided to a solution implementation engine, which iteratively implements solutions to resolve the issue using solution implementation information associated with a given solution. Thus, the data processing engine need not have knowledge regarding the specific hardware and/or software used within the CDN. Similarly, the solution generation engine need not have knowledge of the structure of the CDN and/or configuration of devices associated with the identified issue, such that the solution implementation engine provides a layer of abstraction between a solution and the implementation-specific details used to implement the solution within the CDN.
1. A system comprising:
at least one processor; and
memory, operatively connected to the at least one processor and storing instructions that, when executed by the at least one processor, cause the system to perform a set of operations, the set of operations comprising:
receiving, from a node of a content distribution network (CDN), log data comprising one or more events associated with a computing device of the node;
processing the log data using a model to determine an issue;
selecting a solution generation engine from a plurality of solution generation engines based at least in part on the determined issue;
generating, by the selected solution generating engine and based on the determined issue, a set of solutions to resolve the issue, wherein generating the set of solutions comprises
evaluating a knowledge graph based at least in part on the determined issue to identify a cause node;
evaluating a rule of a relationship to determine whether the determined issue satisfies the rule, wherein the relationship associates a solution node with the cause node; and
based on determining the determined issue satisfies the rule, adding a solution associated with the solution node to the set of solutions;
selecting a first solution of the set of solutions; and
implementing the first solution to resolve the issue.
2. The system of claim 1 , wherein implementing the first solution to resolve the issue comprises:
identifying a set of instructions associated with the first solution; and
performing the identified set of instructions to implement the first solution.
3. The system of claim 1 , wherein the solution generation engine is selected based at least in part on the computing device.
4. The system of claim 1 , wherein the set of operations further comprises:
storing at least a part of the log data associated with the determined issue as training data; and
training an updated model using the training data.
5. The system of claim 4 , wherein the training data further comprises at least a part of the log data associated with a routine operation of the computing device.
6. The system of claim 1 , wherein the set of operations further comprises:
determining the first solution did not resolve the issue;
selecting a second solution of the set of solutions; and
implementing the second solution to resolve the issue.
7. The system of claim 1 , wherein processing the log data comprises:
identifying at least a part of the log data associated with routine operation of the computing device to generate filtered log data; and
processing the filtered log data using the model to determine the issue.
8. The system of claim 1 , wherein the model is a first model and the issue is a first issue, and wherein processing the log data to determine the issue further comprises:
processing the log data using a second model to identify a second issue; and
selecting the first issue as the determined issue based at least in part on determining a performance metric for with the first model is higher than the performance metric for the second model.
9. The system of claim 8 , wherein the performance metric is one of:
a confidence score associated with processing the log data;
an average confidence score based on historical model performance; or
a prediction accuracy.
10. The system of claim 1 , wherein the knowledge graph is generated from at least one of:
a manual;
a knowledge base article;
a trouble ticket;
current event data;
an electronic message;
a planned maintenance data base; or
a network inventory system.
11. The system of claim 1 , wherein the set of operations further comprises:
determining that a performance metric of the model is below a predetermined threshold;
based on determining that the performance metric is below the predetermined threshold, retraining the model using historical log data, wherein the historical log data comprises at least a part of the received log data; and
processing additional log data using the retrained model to determine a second issue.
12. The system of claim 1 , wherein the model is a first model and the set of operations further comprises:
determining that a performance metric of the first model is below a predetermined threshold;
based on determining that the performance metric is below the predetermined threshold, selecting a second model to use in place of the first model; and
processing additional log data using the second model to determine a second issue.
13. The system of claim 1 , wherein selecting the solution generation engine further comprises identifying a hardware device or software package associated with the issue and selecting the solution generation based on the identified hardware device or software package.
14. The system of claim 1 , wherein selecting the solution generation engine further comprises evaluating a computing functionality associated with the issue and selecting the solution generation engine based on the evaluated computing functionality.
15. A method comprising:
receiving, from a node of a content distribution network (CDN), log data comprising one or more events associated with a computing device of the node;
processing the log data using a model to determine an issue;
selecting a solution generation engine from a plurality of solution generation engines based at least in part on the determined issue;
generating, by the selected solution generation engine and based on the determined issue, a set of solutions to resolve the issue, wherein generating the set of solutions comprises evaluating a knowledge graph based at least in part on the determined issue to identify a cause node;
evaluating a rule of a relationship to determine whether the determined issue satisfies the rule, wherein the relationship associates a solution node with the cause node; and
based on determining the determined issue satisfies the rule, adding a solution associated with the solution node to the set of solutions;
selecting a first solution of the set of solutions; and
implementing the first solution to resolve the issue.
16. The method of claim 15 , wherein implementing the first solution to resolve the issue comprises:
identifying a set of instructions associated with the first solution; and
performing the identified set of instructions to implement the first solution.
17. The method of claim 15 , further comprising:
storing at least a part of the log data associated with the determined issue as training data; and
training an updated model using the training data.
18. The method of claim 17 , wherein the training data further comprises at least a part of the log data associated with a routine operation of the computing device.
19. The method of claim 15 , wherein processing the log data comprises:
identifying at least a part of the log data associated with routine operation of the computing device to generate filtered log data; and
processing the filtered log data using the model to determine the issue.