System and method for SLA violation mitigation via multi-level thresholds

Patent metadata

Field	Value
Publication number	US-9843486-B2
Application number	US-201514689040-A
Country	US
Kind code	B2
Filing date	Apr 16, 2015
Priority date	Apr 16, 2015
Publication date	Dec 12, 2017
Grant date	Dec 12, 2017

Abstract

According to one embodiment, a method in a cloud infrastructure for multi-level threshold service level agreement (SLA) violation mitigation. The method includes generating a model for an engineered feature (eF); determining thresholds T 1 , T 2 , and a maximum threshold Tε for the eF based on the model; estimating a value of the eF based on metrics; responsive to determining that the value of the eF exceeds Tε, modifying the values of T 1 and T 2 and modifying the estimation frequency and sending to a cloud orchestrator a message indicating that an SLA violation of type Tε has occurred.

First claim

What is claimed is:

1. A method in a cloud infrastructure of multi-level threshold service level agreement (SLA) violation mitigation, the method comprising:
generating a model for an engineered feature (eF), wherein the eF is based on collected metrics;
determining a first threshold T 1 , a second threshold T 2 , and a maximum threshold Tε for the eF based on the generated model of the eF, wherein the maximum threshold Tε when exceeded indicates that a SLA violation occurred, and wherein the first threshold T 1 and the second threshold T 2 are intermediate thresholds that are smaller than the maximum threshold Tε;
estimating at a predetermined frequency a value of the eF based on real-time metrics;
responsive to determining that the value of the eF exceeds Tε, adjusting the values of T 1 and T 2 and modifying the predetermined frequency and sending to a cloud orchestrator a message indicating that an SLA violation of type Tε has occurred;
responsive to determining that the value of the eF does not exceed Tε, determining whether the value of eF exceeds T 1 ;
responsive to determining that the value of eF exceeds T 1 , determining whether a change in rate of eF (eF′) has been positive;
responsive to determining that the change in rate of eF has been positive, determining whether the value of eF exceeds T 2 ;
responsive to determining that the value of eF exceeds T 2 , determining whether a change in the rate of eF′ (eF″) exceeds a value 1 ;
responsive to determining that the value of eF″ exceeds the value 1 , triggering a SLA violation warning of type value 1 ;
responsive to determining that the value of eF does not exceed T 2 , determining whether the change in the rate of eF″ exceeds a value 2 ;
responsive to determining that the value of eF″ exceeds the value 2 , triggering a SLA violation warning of type value 2 ;
determining one or more elements of the eF that triggered the SLA violation warning; and
sending to the cloud orchestrator a message indicating the one or more elements.

2. The method of claim 1 , wherein the triggering of a SLA violation warning includes sending to the cloud orchestrator a message indicating that a SLA violation is likely to occur soon.

3. The method of claim 2 , wherein the message further includes the type of the SLA violation warning, and wherein the type of the violation warning is one of the type of value 1 and the type of value 2 .

4. The method of claim 1 , wherein the determining the one or more elements of the eF that triggered the SLA violation warning further comprises:
selecting a predetermined number of raw input metrics for the eF that have the highest rate of change; and
selecting a set of raw input metrics used to determine the eF that have exceeded their respective rate of change or rate of rate of change thresholds.

5. The method of claim 1 , wherein the determining the one or more elements of the eF that triggered the SLA violation warning further comprises:
selecting a cluster from a K-means analysis of SLA violation warnings of the eF that the SLA violation warning belongs to, where the K-means analysis of the SLA violation warning is based on collected raw input metrics and eF values of prior SLA violation warnings for that eF.

6. The method of claim 1 , wherein the estimating the value of the eF comprises determining the value of eF using one of a parallel select machine learning model, a parallel-combine machine learning model, and a sequential machine learning model.

7. The method of claim 1 , wherein the generating the model of the eF comprises:
estimating one or more eF values based on historically collected metrics; and
creating a model based on the one or more eF values.

8. The method of claim 1 , wherein the determining the T 1 , T 2 , and the Tε further comprise:
determining the T 1 , T 2 , and the Tε based on at least one of a set of predetermined values, a domain expert, and analysis of past performance.

9. The method of claim 1 , further comprising:
responsive to determining that the value of eF exceeds T 2 , updating a counter value; and
responsive to determining that the counter value exceeds a value x, triggering a SLA violation warning of type value x and resetting the counter value.

10. A non-transitory computer readable medium, having stored thereon a computer program, which when executed by a processor performs the following operations:
generating a model for an engineered feature (eF), wherein the eF is based on collected metrics;
determining a first threshold T 1 , a second threshold T 2 , and a maximum threshold Tε for the eF based on the generated model of the eF, wherein the maximum threshold Tε if exceeded by a value of the eF indicates that a SLA violation occurred, and wherein the first threshold T 1 and the second threshold T 2 are intermediate thresholds that are smaller than the maximum threshold Tε;
estimating at a predetermined frequency a value of the eF based on real-time metrics;
responsive to determining that the value of the eF exceeds Tε, adjusting the values of T 1 and T 2 and modifying the predetermined frequency and sending to a cloud orchestrator a message indicating that an SLA violation of type Tε has occurred;
responsive to determining that the value of the eF does not exceed Tε, determining whether the value of eF exceeds T 1 ;
responsive to determining that the value of eF exceeds T 1 , determining whether a change in rate of eF (eF′) has been positive;
responsive to determining that the change in rate of eF has been positive, determining whether the value of eF exceeds T 2 ;
responsive to determining that the value of eF exceeds T 2 , determining whether a change in the rate of eF′ (eF″) exceeds a value 1 ;
responsive to determining that the value of eF″ exceeds the value 1 , triggering a SLA violation warning of type value 1 ;
responsive to determining that the value of eF does not exceed T 2 , determining whether the change in the rate of eF″ exceeds a value 2 ;
responsive to determining that the value of eF″ exceeds the value 2 , triggering a SLA violation warning of type value 2 ;
determining one or more elements of the eF that triggered the SLA violation warning; and
sending to the cloud orchestrator a message indicating the one or more elements.

11. The non-transitory computer readable medium claim 10 , wherein the triggering of a SLA violation warning includes sending to the cloud orchestrator a message indicating that a SLA violation is likely to occur soon.

12. The non-transitory computer readable medium of claim 10 , wherein the message further includes the type of the SLA violation warning, and wherein the type of the violation warning is one of the type of value 1 and the type of value 2 .

13. The non-transitory computer readable medium of claim 10 , wherein the determining the one or more elements of the eF that triggered the SLA violation warning further comprises:
selecting a predetermined number of raw input metrics for the eF that have the highest rate of change; and
selecting a set of raw input metrics used to determine the eF that have exceeded their respective rate of change or rate of rate of change thresholds.

14. The non-transitory computer readable medium of claim 10 , wherein the determining the one or more elements of the eF that triggered the SLA violation warning further comprises:
selecting a cluster from a K-means analysis of SLA violations warnings of the eF that the SLA violation warning belongs to, where the K-means analysis of the SLA violation warning is based on collected raw input metrics and eF values of prior SLA violation warnings for that

Assignees

• Ericsson Telefon Ab L M
• ERICSSON TELEFON AB L M (publ)

Inventors

• Feller Eugen
• Forgeat Julien
• Shirazipour Meral
• Xia Ming

Classifications

• H04L41/0681

  Configuration of triggering conditions · CPC title

• H04L41/5025

  by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade · CPC title

• H04L43/0817

  by checking functioning · CPC title

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• H04L41/5009 Primary

  Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF] · CPC title