US 2022/0019585 A1 · TERM VECTOR MODELING OF DATABASE WORKLOADS

Abstract

Techniques for managing database workloads using similarity measures based on queries executed are described. Classical techniques from information retrieval are applied to the domain of database workload management. Specifically, the technique of using document term vectors to compute similarity measures are applied using the conceptual mapping of SQL workloads as “documents” composed of SQL queries as “terms.” The techniques include generating two or more sets of workloads with each workload representing a set of queries executed on at least one database. Based on the sets of workloads, workload term vectors are calculated that represent the set of queries executed on the database. Then, based on the calculated workload vectors, a similarity score is generated between the two or more sets of workloads.

Claims20

Dependent claims

A method comprising:

generating a first set of one or more workloads representing a first set of queries of a plurality of queries executed on at least one database;
generating a first plurality of query vectors corresponding respectively to each of the first set of queries;
calculating, based on the first plurality of query vectors, a first workload vector corresponding to the first set of workloads;
generating a second set of one or more workloads representing a second set of queries of the plurality of queries;
generating a second plurality of query vectors corresponding respectively to each of the second set of queries;
calculating, based on the second plurality of query vectors, a second workload vector corresponding to the second set of workloads;
and
generating a similarity score between the first set of workloads and the second set of workloads based on the first workload vector and the second workload vector,
wherein the method is performed by at least one device including a hardware processor.

6 more elements

A system, comprising:

one or more hardware processors;
a non-transitory computer readable medium comprising instructions which, when executed by the one or more hardware processors, causes performance of operations comprising:
generating a first set of one or more workloads representing a first set of queries of a plurality of queries;
generating a first plurality of query vectors corresponding respectively to each of the first set of queries;
calculating, based on the first plurality of query vectors, a first workload vector corresponding to the first set of workloads;
generating a second set of one or more workloads representing a second set of queries of the plurality of queries;
generating a second plurality of query vectors corresponding respectively to each of the second set of queries;
calculating, based on the second plurality of query vectors, a second workload vector corresponding to the second set of workloads;
and
generating a similarity score between the first set of workloads and the second set of workloads based on the first workload vector and the second workload vector.

7 more elements

A non-transitory computer readable medium comprising instructions which, when executed by one or more hardware processors, causes performance of operations comprising:

generating a first set of one or more workloads representing a first set of queries of a plurality of queries;
generating a first plurality of query vectors corresponding respectively to each of the first set of queries;
calculating, based on the first plurality of query vectors, a first workload vector corresponding to the first set of workloads;
generating a second set of one or more workloads representing a second set of queries of the plurality of queries;
generating a second plurality of query vectors corresponding respectively to each of the second set of queries;
calculating, based on the second plurality of query vectors, a second workload vector corresponding to the second set of workloads;
and
generating a similarity score between the first set of workloads and the second set of workloads based on the first workload vector and the second workload vector.

5 more elements

Dependencies are inferred from the claim text.

Term vector modeling of database workloads

Life of this patent

Term vector modeling of database workloads

Life of this patent

Abstract

Claims20