Graph database query classification based on previous queries stored in repository
Examples for graph database query classification include receiving a graph query and determining if the graph query matches benchmark data. In the event that the graph query does not match benchmark data, the query may be parsed, a canonical internal representation of the query may be determined, the representation may be mapped to a rule, and the query may be classified based on the rule. In the event that the confidence score for the query classification does not exceed a threshold, the query may be sent to a synthetic graph or synopsis for simulation. In some examples, the simulation may include selecting computationally expensive graph operators in the query for simulation.
1. A graph database query classification system, comprising:
a repository of benchmark data;
a graph processor engine; and
a memory storing instructions that when executed cause the graph processor engine to:
receive a graph query from an application;
determine whether the graph query matches a previously submitted query stored in the repository of benchmark data;
in response to a determination that the graph query matches a previously submitted query stored in the repository of benchmark data, classify the graph query to a classification associated with the previously submitted query;
in response to a determination that the graph query does not match a previously submitted query stored in the repository of benchmark data, determine a canonical internal representation of the graph query, map the canonical internal representation of the graph query to a rule, and classify the graph query to a classification associated with the rule;
determine a confidence score of the classification of the graph query;
in response to a determination that the confidence score of the classification of the graph query is less than a threshold score, determine that the classification of the graph query does not meet a predetermined condition; and
in response to a determination that the classification of the graph query does not meet the predetermined condition, simulate an execution of the graph query on a synthetic graph.
2. The graph database query classification system of claim 1 , wherein the rule is determined based on historical logs of queries stored in the repository of benchmark data.
3. The graph database query classification system of claim 1 , wherein, to simulate the execution of the graph query, the instructions are executable to cause the graph processor engine to:
select an operator in the graph query; and
simulate the execution of the graph query on the synthetic graph for the selected operator.
4. The graph database query classification system of claim 3 , wherein the instructions are executable to cause the graph processor engine to:
determine whether a total number of graph nodes accessed during the simulated execution of the graph query is greater than a threshold number of nodes; and
in response to a determination that the total number of graph nodes accessed in the simulated execution of the graph query is greater than the threshold number of nodes, classify the graph query as a graph analytic query.
5. The graph database query classification system of claim 4 , wherein the instructions are executable to cause the graph processor engine to:
in response to a determination that the total number of graph nodes accessed in the simulated execution of the graph query is less than the threshold number of nodes, classify the graph query as a graph navigation query.
6. The graph database query classification system of claim 3 , wherein the selected operator is an operator that demands more computing resources than other operators in the graph query.
7. A method for classifying a graph database query, comprising:
receiving a graph query;
determining whether the graph query matches a previously submitted query stored in a repository of benchmark data;
in response to a determination that the graph query matches a previously submitted query stored in the repository of benchmark data, classifying the graph query to a classification associated with the previously submitted query;
in response to a determination that the graph query does not match a previously submitted query stored in the repository of benchmark data, parsing the graph query, determining a canonical internal representation of the graph query, mapping the canonical internal representation of the graph query to a rule, and classifying the graph query into a classification associated with the rule;
determining a confidence score for the classification of the graph query; and
in response to a determination that the confidence score does not exceed a threshold score, sending the graph query to a synthetic graph for simulation.
8. The method of claim 7 , further comprising:
selecting operators in the graph query; and
simulating an execution of the graph query on the synthetic graph for the selected operators.
9. The method of claim 8 , further comprising:
determining whether a total number of graph nodes accessed in the simulated execution of the graph query is greater than a threshold number of nodes; and
in response to a determination that the total number of graph nodes accessed in the simulated execution is greater than the threshold number of nodes, classifying the graph query as a graph analytic query.
10. The method of claim 9 , further comprising:
in response to a determination that the total number of graph nodes accessed in the simulated execution is less than the threshold number of nodes, classifying the graph query as a graph navigation query.
11. The method of claim 8 , wherein the selected operators are operators that demand more computing resources than other operators in the graph query.
12. An article comprising at least one non-transitory machine-readable storage medium that stores instructions executable by a processing resource of a graph database classification system to:
determine whether a received query matches a previously submitted query stored in a repository of benchmark data;
in response to a determination that the received query matches a previously submitted query stored in the repository of benchmark data, classify the received query to a classification associated with the previously submitted query;
in response to a determination that the received query does not match a previously submitted query stored in the repository of benchmark data, determine a canonical internal representation of the received query, map the canonical internal representation of the received query to a rule, and classify the received query to a classification associated with the rule;
determine a confidence score of the classification of the received query; and
in response to a determination that the confidence score is less than a threshold score, simulate an execution of the received query on a synthetic graph.
13. The article of claim 12 , wherein, to simulate the execution of the received query, the instructions are executable to cause the processing resource to:
select an operator in the received query; and
simulate the execution of the received query with the selected operator.
14. The article of claim 13 , wherein the instructions are executable to cause the processing resource to:
determine whether a total number of graph elements accessed in the simulated execution is greater than a threshold number; and
in response to a determination that the total number of graph elements accessed in the simulated execution is greater than the threshold number, classify the received query as a graph analytic query.
15. The article of claim 14 , wherein the instructions are executable to cause the processing resource to:
in response to a determination that the total number of graph elements is less than the threshold number, classify the received query as a graph navigation query.
16. The article of claim 15 , wherein the graph elements are a plurality of graph edges or a plurality of graph nodes.