Performance modeling for cloud applications

1. A method for performance modeling of a plurality of microservices , the method comprising:

deploying the plurality of microservices within a network, the plurality of microservices communicatively coupled to generate at least one service chain for providing at least one service;

based on a resource allocation configuration, determining an initial set of training data for the plurality of microservices within the network;

excluding at least a portion of data from the initial set of training data to generate a subset of training data; and

generating a Quality of Service (QoS) behaviour model based on the subset of the training data, wherein selecting the QoS behaviour model based on the set of training data, comprises:

testing a plurality of hypothesis functions for the plurality of microservices in a service chain;

selecting one of a plurality of hypothesis functions that has a highest goodness of fit and cross-validation results to the set of training data; and

training the QoS behaviour model to estimate at least one QoS metric.

2. The method of claim 1 , wherein excluding the portion of data from the initial set of training data to generate the subset of training data comprises:

isolating resources of the plurality of microservices;

selecting one of the plurality of microservices;

assigning a maximum respective resource allocation configuration to each of a plurality of resources associated with the plurality of microservices;

determining a saturation point for the one of the plurality of microservices; and

excluding a saturation area associated with the saturation point from the initial set of training data when generating the subset of training data.

3. The method of claim 2 , wherein the saturation point comprises a point when a quality of service associated with the one of the plurality of microservices starts saturating while increasing resources.

4. A non-transitory computer-readable medium comprising computer program, the computer program comprising instructions which when executed on a computer perform the method of claim 1 .

5. A network node for performance modeling of a plurality of microservices, the network node comprising:

processing circuitry configured to:

deploy the plurality of microservices within a network, the plurality of microservices communicatively coupling to generate at least one service chain for providing at least one service;

based on a resource allocation configuration, determine an initial set of training data for the plurality of microservices within the network;

exclude at least a portion of data from the initial set of training data to generate a subset of training data; and

generate a Quality of Service (QoS) behaviour model based on the subset of the training data, wherein when selecting the QoS behaviour model based on the set of training data the processing circuitry is configured to:

test a plurality of hypothesis functions for the plurality of microservices in a service chain;

select one of a plurality of hypothesis functions that has a highest goodness of fit and cross-validation results to the set of training data; and

train the QoS behaviour model to estimate at least one QoS metric.

6. The network node of claim 5 , wherein when excluding the portion of data from the initial set of training data to generate the subset of training data the processing circuitry is configured to:

isolate resources of the plurality of microservices;

select one of the plurality of microservices;

assign a maximum respective resource allocation configuration to each of a plurality of resources associated with the plurality of microservices;

determine a saturation point for the one of the plurality of microservices; and

exclude a saturation area associated with the saturation point from the initial set of training data when generating the subset of training data.

7. The network node of claim 6 , wherein the saturation point comprises a point when a quality of service associated with the one of the plurality of microservices starts saturating while increasing resources.

8. The network node of claim 6 , wherein the processing circuitry is configured to repeat the steps of isolating, selecting, assigning, determining, and excluding for a randomly selected subset of the plurality of microservices.

9. The network node of claim 5 , wherein the network comprises a lab environment, the lab environment comprising at least one infrastructure element, the at least one infrastructure element of the lab environment selected to emulate a production environment to minimize prediction errors.

10. The network node of claim 9 , wherein a network topology of the lab environment emulates a network topology of the production environment.

11. The network node of claim 9 , wherein the at least one infrastructure element comprises at least one of equipment, hardware, operating system, and bandwidth selected to emulate the production environment.

12. The network node of claim 5 , wherein when deploying the plurality of microservices within the network the processing circuitry is configured to determine at least one network node to host each of the plurality of microservices.

13. The network node of claim 12 , wherein a communication capacity of each of the plurality of microservices is considered when determining the at least one network node to host each of the plurality of microservices.

14. The network node of claim 13 , wherein the plurality of microservices are hosted on a plurality of network nodes, each of the plurality of network nodes being selected to host at least one of the plurality of microservices based on at least one affinity rule or randomly.

15. The network node of claim 5 , wherein when deploying the plurality of microservices within the network the processing circuitry is configured to minimize factors affecting QoS such that the resource allocation configuration is the only deterrent to getting a higher QoS value.

16. The network node of claim 5 , wherein the processing circuitry is configured to use QoS behaviour models of the plurality of microservices to determine a value of the at least one QoS metric for the plurality of microservices in the service chain.

17. The network node of claim 5 , wherein the processing circuitry is configured to:

use QoS behaviour models of the plurality of microservices to determine an optimal amount of resource capacities of each of the plurality of microservices belonging to the at least one service chain.

18. The network node of claim 17 , wherein the optimal amount of resource capacities of each of the plurality of microservices is determined based on at least one QoS constraint for a user.

19. The network node of claim 5 , wherein the QoS behaviour model of the plurality of microservices comprises a machine-learning model.