continuity.service.cobra

Cobra

The Cobra service prepares the data required for generating load tests. Users can upload data and later refer to them.

Uploading Data

The data can be uploaded via the REST endpoint POST /measurement-data/{app-id}/{version}/{type}. The app-id and version identify the application to which the data belongs. The type can be open-xtrace (see OPEN.xtrace), access-logs (Apache common log format), csv (CSV file with named columns), or session-logs (see here for an example).

Cobra automatically learns workload models and intensities (numbers of concurrent users) from the data, which it stores into an Elasticsearch. For the clustering, it utilizes the Clustinator service.

Artifact Creation

When being triggered, Cobra can generate the artifacts traces, sessions, and behavior-model. In addition, it always sets the intensity artifact. For the intensity calculation (which can include time series forecasting), the Forecastic service is used.

Configuration

The service can be configured using the configuration management of the Orchestrator. The following example illustrates the configuration space. We recommend to use the CLI for creating and editing the configurations.

---
service: cobra
app-id: sis
time-zone: Europe/Berlin
traces:
  retention: P14D
  map-to-idpa: true
  discard-unmapped: true
  log-unmapped: false
  stop-on-failure: true
tailoring:
- - all
sessions:
  timeout: PT30M
  omit: false
  hash-id: false
  ignore-redirects: true
clustering:
  interval: P7D
  overlap: P21D
  lookback: 200
  initial:
    strategy: kmeans
    k: 20
    parallelize: -2
    num-seedings: 30
    convergence-tolerance: 1.0E-5
    quantile-range: 0.95
  append:
    strategy: minimum-distance
    min-sample-size: 500
    radius-factor: 1.1
  omit: false
intensity:
  resolution: PT1M
context:
  auto-detect: true
  variables:
    black_friday:
      type: boolean
      ignore-by-default: false

Publications

The tailoring field in the configuration relates to log-based service-tailoring, whis is introduced in the following publication:

Henning Schulz, Tobias Angerstein, Dušan Okanović, and André van Hoorn: Microservice-tailored generation of session-based workload models for representative load testing (full paper), Proceedings of the 27th IEEE International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS 2019), 2019