Framework for estimation of resource usage and execution time of workloads in a heterogeneous infrastructure

1 . A method for performing operations concerning a model that is operable to predict resource usage and an execution time of computing workloads, the operations comprising:

extracting a fingerprint associated with telemetry data, wherein the telemetry data was generated based on performance of computing resources of one of the computing workloads, in a constrained infrastructure;

checking a fingerprint catalog to determine if there is a same or similar fingerprint to the extracted fingerprint;

when the same or similar fingerprint is found in the fingerprint catalog,

mapping the telemetry data of the constrained infrastructure to telemetry data of a target infrastructure based on the same or similar fingerprint in the model; and

predicting an execution time for the computing workload in the target infrastructure based on an executing time in an infrastructure, which has the same or similar fingerprint in the fingerprint catalog;

when the same or similar fingerprint is not found in the fingerprint catalog,

inserting the extracted fingerprint into the fingerprint catalog, and

generating a retrained model by retraining the model using the telemetry data associated with the extracted fingerprint; and

dynamically allocating computing resources in the constrained infrastructure to the one of the computing workloads based on the telemetry data of the target infrastructure and the retrained model.

2 . The method as recited in claim 1 , wherein the model comprises a deep neural network.

3 . The method as recited in claim 1 , wherein running the model for one of the computing workloads generates a predicted resource usage by the workload, and a predicted execution time for that computing workload.

4 . The method as recited in claim 1 , further comprising running the retrained model with the telemetry data that is associated with the fingerprint, and with telemetry data obtained by executing the computing workload in the target infrastructure.

5 . The method as recited in claim 1 , wherein the model uses inputs to generate predicted resource usage and predicted execution time for a computing workload, and the inputs comprise encoded data.

6 . The method as recited in claim 5 , wherein the encoded data is in a form of an input vector, and the model multiplies the input vector with a matrix of weights, where each weight corresponds to a respective input of the input vector.

7 . The method as recited in claim 1 , wherein the model maps the telemetry data corresponding to the constrained infrastructure to a combination that comprises the telemetry data corresponding to the target infrastructure and an execution time of a computing workload in the target infrastructure.

8 . The method as recited in claim 1 , wherein the model generates execution time and resource usage predictions without running the computing workload in the constrained infrastructure.

9 . The method as recited in claim 1 , wherein the telemetry data comprises one or both of CPU usage and RAM usage.

10 . The method as recited in claim 1 , wherein the telemetry data is represented as an aligned multi-channel time series.

11 . A computer readable storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations concerning a model that is operable to predict resource usage and an execution time of computing workloads, the operations comprising:

extracting a fingerprint associated with telemetry data, wherein the telemetry data was generated based on performance of one of the computing workloads, in a constrained infrastructure;

checking a fingerprint catalog to determine if there is a same or similar fingerprint to the extracted fingerprint;

when the same or similar fingerprint is found in the fingerprint catalog,

mapping the telemetry data of the constrained infrastructure to telemetry data of a target infrastructure based on the same or similar fingerprint in the model; and

predicting an execution time for the computing workload in the target infrastructure based on an executing time in an infrastructure, which has the same or similar fingerprint in the fingerprint catalog;

when the same or similar fingerprint is not found in the fingerprint catalog,

inserting the extracted fingerprint into the fingerprint catalog, and

generating a retrained model by retraining the model using the telemetry data associated with the extracted fingerprint; and

dynamically allocating computing resources in the constrained infrastructure to the one of the computing workloads based on the telemetry data of the target infrastructure and the retrained model.

12 . The computer readable storage medium as recited in claim 11 , wherein the model comprises a deep neural network.

13 . The computer readable storage medium as recited in claim 11 , wherein running the model for one of the computing workloads generates a predicted resource usage by the workload, and a predicted execution time for that computing workload.

14 . The computer readable storage medium as recited in claim 11 , further comprising running the retrained model with the telemetry data that is associated with the fingerprint, and with telemetry data obtained by executing the computing workload in the target infrastructure.

15 . The computer readable storage medium as recited in claim 11 , wherein the model uses inputs to generate predicted resource usage and predicted execution time for a computing workload, and the inputs comprise encoded data.

16 . The computer readable storage medium as recited in claim 15 , wherein the encoded data is in a form of an input vector, and the model multiplies the input vector with a matrix of weights, where each weight corresponds to a respective input of the input vector.

17 . The computer readable storage medium as recited in claim 11 , wherein the model maps the telemetry data corresponding to the constrained infrastructure to a combination that comprises the telemetry data corresponding to the target infrastructure and an execution time of a computing workload in the target infrastructure.

18 . The computer readable storage medium as recited in claim 11 , wherein the model generates execution time and resource usage predictions without running the computing workload in the constrained infrastructure.

19 . The computer readable storage medium as recited in claim 11 , wherein the telemetry data comprises one or both of CPU usage and RAM usage.

20 . The computer readable storage medium as recited in claim 11 , wherein the telemetry data is represented as an aligned multi-channel time series.